Modeling In H2O

Supervised

H2OANOVAGLMEstimator

class h2o.estimators.anovaglm.H2OANOVAGLMEstimator(model_id=None, training_frame=None, seed=-1, response_column=None, ignored_columns=None, ignore_const_cols=True, score_each_iteration=False, offset_column=None, weights_column=None, family='auto', tweedie_variance_power=0.0, tweedie_link_power=1.0, theta=0.0, solver='irlsm', missing_values_handling='mean_imputation', plug_values=None, compute_p_values=True, standardize=True, non_negative=False, max_iterations=0, link='family_default', prior=0.0, alpha=None, lambda_=[0.0], lambda_search=False, stopping_rounds=0, stopping_metric='auto', early_stopping=False, stopping_tolerance=0.001, balance_classes=False, class_sampling_factors=None, max_after_balance_size=5.0, max_runtime_secs=0.0, save_transformed_framekeys=False, highest_interaction_term=0, nparallelism=4, type=0)[source]

Bases: h2o.estimators.estimator_base.H2OEstimator

ANOVA for Generalized Linear Model

H2O ANOVAGLM is used to calculate Type III SS which is used to evaluate the contributions of individual predictors and their interactions to a model. Predictors or interactions with negligible contributions to the model will have high p-values while those with more contributions will have low p-values.

property Lambda

DEPRECATED. Use self.lambda_ instead

property alpha

Distribution of regularization between the L1 (Lasso) and L2 (Ridge) penalties. A value of 1 for alpha represents Lasso regression, a value of 0 produces Ridge regression, and anything in between specifies the amount of mixing between the two. Default value of alpha is 0 when SOLVER = ‘L-BFGS’; 0.5 otherwise.

Type: List[float].

property balance_classes

Balance training data class counts via over/under-sampling (for imbalanced data).

Type: bool, defaults to False.

property class_sampling_factors

Desired over/under-sampling ratios per class (in lexicographic order). If not specified, sampling factors will be automatically computed to obtain class balance during training. Requires balance_classes.

Type: List[float].

property compute_p_values

Request p-values computation, p-values work only with IRLSM solver and no regularization

Type: bool, defaults to True.

property early_stopping

Stop early when there is no more relative improvement on train or validation (if provided).

Type: bool, defaults to False.

property family

Family. Use binomial for classification with logistic regression, others are for regression problems.

Type: Literal["auto", "gaussian", "binomial", "fractionalbinomial", "quasibinomial", "poisson", "gamma", "tweedie", "negativebinomial"], defaults to "auto".

property highest_interaction_term

Limit the number of interaction terms, if 2 means interaction between 2 columns only, 3 for three columns and so on… Default to 2.

Type: int, defaults to 0.

property ignore_const_cols

Ignore constant columns.

Type: bool, defaults to True.

property ignored_columns

Names of columns to ignore for training.

Type: List[str].

property lambda_

Regularization strength

Type: List[float], defaults to [0.0].

Use lambda search starting at lambda max, given lambda is then interpreted as lambda min

Type: bool, defaults to False.

Link function.

Type: Literal["family_default", "identity", "logit", "log", "inverse", "tweedie", "ologit"], defaults to "family_default".

property max_after_balance_size

Maximum relative size of the training data after balancing class counts (can be less than 1.0). Requires balance_classes.

Type: float, defaults to 5.0.

property max_iterations

Maximum number of iterations

Type: int, defaults to 0.

property max_runtime_secs

Maximum allowed runtime in seconds for model training. Use 0 to disable.

Type: float, defaults to 0.0.

property missing_values_handling

Handling of missing values. Either MeanImputation, Skip or PlugValues.

Type: Literal["mean_imputation", "skip", "plug_values"], defaults to "mean_imputation".

property non_negative

Restrict coefficients (not intercept) to be non-negative

Type: bool, defaults to False.

property nparallelism

Number of models to build in parallel. Default to 4. Adjust according to your system.

Type: int, defaults to 4.

property offset_column

Offset column. This will be added to the combination of columns before applying the link function.

Type: str.

property plug_values

Plug Values (a single row frame containing values that will be used to impute missing values of the training/validation frame, use with conjunction missing_values_handling = PlugValues)

Type: Union[None, str, H2OFrame].

property prior

Prior probability for y==1. To be used only for logistic regression iff the data has been sampled and the mean of response does not reflect reality.

Type: float, defaults to 0.0.

property response_column

Response variable column.

Type: str.

result()[source]

Get result frame that contains information about the model building process like for modelselection and anovaglm. :return: the H2OFrame that contains information about the model building process like for modelselection and anovaglm.

property save_transformed_framekeys

true to save the keys of transformed predictors and interaction column.

Type: bool, defaults to False.

property score_each_iteration

Whether to score during each iteration of model training.

Type: bool, defaults to False.

property seed

Seed for pseudo random number generator (if applicable)

Type: int, defaults to -1.

property solver

AUTO will set the solver based on given data and the other parameters. IRLSM is fast on on problems with small number of predictors and for lambda-search with L1 penalty, L_BFGS scales better for datasets with many columns.

Type: Literal["auto", "irlsm", "l_bfgs", "coordinate_descent_naive", "coordinate_descent", "gradient_descent_lh", "gradient_descent_sqerr"], defaults to "irlsm".

property standardize

Standardize numeric columns to have zero mean and unit variance

Type: bool, defaults to True.

property stopping_metric

Metric to use for early stopping (AUTO: logloss for classification, deviance for regression and anomaly_score for Isolation Forest). Note that custom and custom_increasing can only be used in GBM and DRF with the Python client.

Type: Literal["auto", "deviance", "logloss", "mse", "rmse", "mae", "rmsle", "auc", "aucpr", "lift_top_group", "misclassification", "mean_per_class_error", "custom", "custom_increasing"], defaults to "auto".

property stopping_rounds

Early stopping based on convergence of stopping_metric. Stop if simple moving average of length k of the stopping_metric does not improve for k:=stopping_rounds scoring events (0 to disable)

Type: int, defaults to 0.

property stopping_tolerance

Relative tolerance for metric-based stopping criterion (stop if relative improvement is not at least this much)

Type: float, defaults to 0.001.

property theta

Theta

Type: float, defaults to 0.0.

property training_frame

Id of the training data frame.

Type: Union[None, str, H2OFrame].

Tweedie link power

Type: float, defaults to 1.0.

property tweedie_variance_power

Tweedie variance power

Type: float, defaults to 0.0.

property type

Refer to the SS type 1, 2, 3, or 4. We are currently only supporting 3

Type: int, defaults to 0.

property weights_column

Column with observation weights. Giving some observation a weight of zero is equivalent to excluding it from the dataset; giving an observation a relative weight of 2 is equivalent to repeating that row twice. Negative weights are not allowed. Note: Weights are per-row observation weights and do not increase the size of the data frame. This is typically the number of times a row is repeated, but non-integer values are supported as well. During training, rows with higher weights matter more, due to the larger loss function pre-factor. If you set weight = 0 for a row, the returned prediction frame at that row is zero and this is incorrect. To get an accurate prediction, remove all rows with weight == 0.

Type: str.

H2OCoxProportionalHazardsEstimator

class h2o.estimators.coxph.H2OCoxProportionalHazardsEstimator(model_id=None, training_frame=None, start_column=None, stop_column=None, response_column=None, ignored_columns=None, weights_column=None, offset_column=None, stratify_by=None, ties='efron', init=0.0, lre_min=9.0, max_iterations=20, interactions=None, interaction_pairs=None, interactions_only=None, use_all_factor_levels=False, export_checkpoints_dir=None, single_node_mode=False)[source]

Bases: h2o.estimators.estimator_base.H2OEstimator

Cox Proportional Hazards

Trains a Cox Proportional Hazards Model (CoxPH) on an H2O dataset.

property export_checkpoints_dir

Automatically export generated models to this directory.

Type: str.

Examples

>>> import tempfile
>>> from os import listdir
>>> heart = h2o.import_file("http://s3.amazonaws.com/h2o-public-test-data/smalldata/coxph_test/heart.csv")
>>> predictor = "age"
>>> response = "event"
>>> checkpoints_dir = tempfile.mkdtemp()
>>> coxph = H2OCoxProportionalHazardsEstimator(start_column="start",
...                                            stop_column="stop",
...                                            export_checkpoints_dir=checkpoints_dir)
>>> coxph.train(x=predictor,
...             y=response,
...             training_frame=heart)
>>> len(listdir(checkpoints_dir))
property ignored_columns

Names of columns to ignore for training.

Type: List[str].

property init

Coefficient starting value.

Type: float, defaults to 0.0.

Examples

>>> heart = h2o.import_file("http://s3.amazonaws.com/h2o-public-test-data/smalldata/coxph_test/heart.csv")
>>> predictor = "age"
>>> response = "event"
>>> heart_coxph = H2OCoxProportionalHazardsEstimator(start_column="start",
...                                                  stop_column="stop",
...                                                  init=2.9)
>>> heart_coxph.train(x=predictor,
...                   y=response,
...                   training_frame=heart)
>>> heart_coxph.scoring_history()
property interaction_pairs

A list of pairwise (first order) column interactions.

Type: List[tuple].

Examples

>>> heart = h2o.import_file("http://s3.amazonaws.com/h2o-public-test-data/smalldata/coxph_test/heart.csv")
>>> predictor = "age"
>>> response = "event"
>>> interaction_pairs = [("start","stop")]
>>> heart_coxph = H2OCoxProportionalHazardsEstimator(start_column="start",
...                                                  stop_column="stop",
...                                                  interaction_pairs=interaction_pairs)
>>> heart_coxph.train(x=predictor,
...                   y=response,
...                   training_frame=heart)
>>> heart_coxph.scoring_history()
property interactions

A list of predictor column indices to interact. All pairwise combinations will be computed for the list.

Type: List[str].

Examples

>>> heart = h2o.import_file("http://s3.amazonaws.com/h2o-public-test-data/smalldata/coxph_test/heart.csv")
>>> predictor = "age"
>>> response = "event"
>>> interactions = ['start','stop']
>>> heart_coxph = H2OCoxProportionalHazardsEstimator(start_column="start",
...                                                  stop_column="stop",
...                                                  interactions=interactions)
>>> heart_coxph.train(x=predictor,
...                   y=response,
...                   training_frame=heart)
>>> heart_coxph.scoring_history()
property interactions_only

A list of columns that should only be used to create interactions but should not itself participate in model training.

Type: List[str].

Examples

>>> heart = h2o.import_file("http://s3.amazonaws.com/h2o-public-test-data/smalldata/coxph_test/heart.csv")
>>> predictor = "age"
>>> response = "event"
>>> interactions = ['start','stop']
>>> heart_coxph = H2OCoxProportionalHazardsEstimator(start_column="start",
...                                                  stop_column="stop",
...                                                  interactions_only=interactions)
>>> heart_coxph.train(x=predictor,
...                   y=response,
...                   training_frame=heart)
>>> heart_coxph.scoring_history()
property lre_min

Minimum log-relative error.

Type: float, defaults to 9.0.

Examples

>>> heart = h2o.import_file("http://s3.amazonaws.com/h2o-public-test-data/smalldata/coxph_test/heart.csv")
>>> predictor = "age"
>>> response = "event"
>>> heart_coxph = H2OCoxProportionalHazardsEstimator(start_column="start",
...                                                  stop_column="stop",
...                                                  lre_min=5)
>>> heart_coxph.train(x=predictor,
...                   y=response,
...                   training_frame=heart)
>>> heart_coxph.scoring_history()
property max_iterations

Maximum number of iterations.

Type: int, defaults to 20.

Examples

>>> heart = h2o.import_file("http://s3.amazonaws.com/h2o-public-test-data/smalldata/coxph_test/heart.csv")
>>> predictor = "age"
>>> response = "event"
>>> heart_coxph = H2OCoxProportionalHazardsEstimator(start_column="start",
...                                                  stop_column="stop",
...                                                  max_iterations=50)
>>> heart_coxph.train(x=predictor,
...                   y=response,
...                   training_frame=heart)
>>> heart_coxph.scoring_history()
property offset_column

Offset column. This will be added to the combination of columns before applying the link function.

Type: str.

Examples

>>> heart = h2o.import_file("http://s3.amazonaws.com/h2o-public-test-data/smalldata/coxph_test/heart.csv")
>>> predictor = "age"
>>> response = "event"
>>> heart_coxph = H2OCoxProportionalHazardsEstimator(start_column="start",
...                                                  stop_column="stop",
...                                                  offset_column="transplant")
>>> heart_coxph.train(x=predictor,
...                   y=response,
...                   training_frame=heart)
>>> heart_coxph.scoring_history()
property response_column

Response variable column.

Type: str.

property single_node_mode

Run on a single node to reduce the effect of network overhead (for smaller datasets)

Type: bool, defaults to False.

property start_column

Start Time Column.

Type: str.

Examples

>>> heart = h2o.import_file("http://s3.amazonaws.com/h2o-public-test-data/smalldata/coxph_test/heart.csv")
>>> predictor = "age"
>>> response = "event"
>>> train, valid = heart.split_frame(ratios=[.8])
>>> heart_coxph = H2OCoxProportionalHazardsEstimator(start_column="start",
...                                                  stop_column="stop")
>>> heart_coxph.train(x=predictor,
...                   y=response,
...                   training_frame=train,
...                   validation_frame=valid)
>>> heart_coxph.scoring_history()
property stop_column

Stop Time Column.

Type: str.

Examples

>>> heart = h2o.import_file("http://s3.amazonaws.com/h2o-public-test-data/smalldata/coxph_test/heart.csv")
>>> predictor = "age"
>>> response = "event"
>>> train, valid = heart.split_frame(ratios=[.8])
>>> heart_coxph = H2OCoxProportionalHazardsEstimator(start_column="start",
...                                                  stop_column="stop")
>>> heart_coxph.train(x=predictor,
...                   y=response,
...                   training_frame=train,
...                   validation_frame=valid)
>>> heart_coxph.scoring_history()
property stratify_by

List of columns to use for stratification.

Type: List[str].

property ties

Method for Handling Ties.

Type: Literal["efron", "breslow"], defaults to "efron".

Examples

>>> heart = h2o.import_file("http://s3.amazonaws.com/h2o-public-test-data/smalldata/coxph_test/heart.csv")
>>> predictor = "age"
>>> response = "event"
>>> train, valid = heart.split_frame(ratios=[.8])
>>> heart_coxph = H2OCoxProportionalHazardsEstimator(start_column="start",
...                                                  stop_column="stop",
...                                                  ties="breslow")
>>> heart_coxph.train(x=predictor,
...                   y=response,
...                   training_frame=train,
...                   validation_frame=valid)
>>> heart_coxph.scoring_history()
property training_frame

Id of the training data frame.

Type: Union[None, str, H2OFrame].

Examples

>>> heart = h2o.import_file("http://s3.amazonaws.com/h2o-public-test-data/smalldata/coxph_test/heart.csv")
>>> predictor = "age"
>>> response = "event"
>>> train, valid = heart.split_frame(ratios=[.8])
>>> heart_coxph = H2OCoxProportionalHazardsEstimator(start_column="start",
...                                                  stop_column="stop")
>>> heart_coxph.train(x=predictor,
...                   y=response,
...                   training_frame=train,
...                   validation_frame=valid)
>>> heart_coxph.scoring_history()
property use_all_factor_levels

(Internal. For development only!) Indicates whether to use all factor levels.

Type: bool, defaults to False.

Examples

>>> heart = h2o.import_file("http://s3.amazonaws.com/h2o-public-test-data/smalldata/coxph_test/heart.csv")
>>> predictor = "age"
>>> response = "event"
>>> heart_coxph = H2OCoxProportionalHazardsEstimator(start_column="start",
...                                                  stop_column="stop",
...                                                  use_all_factor_levels=True)
>>> heart_coxph.train(x=predictor,
...                   y=response,
...                   training_frame=heart)
>>> heart_coxph.scoring_history()
property weights_column

Column with observation weights. Giving some observation a weight of zero is equivalent to excluding it from the dataset; giving an observation a relative weight of 2 is equivalent to repeating that row twice. Negative weights are not allowed. Note: Weights are per-row observation weights and do not increase the size of the data frame. This is typically the number of times a row is repeated, but non-integer values are supported as well. During training, rows with higher weights matter more, due to the larger loss function pre-factor. If you set weight = 0 for a row, the returned prediction frame at that row is zero and this is incorrect. To get an accurate prediction, remove all rows with weight == 0.

Type: str.

H2ODeepLearningEstimator

class h2o.estimators.deeplearning.H2ODeepLearningEstimator(model_id=None, training_frame=None, validation_frame=None, nfolds=0, keep_cross_validation_models=True, keep_cross_validation_predictions=False, keep_cross_validation_fold_assignment=False, fold_assignment='auto', fold_column=None, response_column=None, ignored_columns=None, ignore_const_cols=True, score_each_iteration=False, weights_column=None, offset_column=None, balance_classes=False, class_sampling_factors=None, max_after_balance_size=5.0, max_confusion_matrix_size=20, checkpoint=None, pretrained_autoencoder=None, overwrite_with_best_model=True, use_all_factor_levels=True, standardize=True, activation='rectifier', hidden=[200, 200], epochs=10.0, train_samples_per_iteration=-2, target_ratio_comm_to_comp=0.05, seed=-1, adaptive_rate=True, rho=0.99, epsilon=1e-08, rate=0.005, rate_annealing=1e-06, rate_decay=1.0, momentum_start=0.0, momentum_ramp=1000000.0, momentum_stable=0.0, nesterov_accelerated_gradient=True, input_dropout_ratio=0.0, hidden_dropout_ratios=None, l1=0.0, l2=0.0, max_w2=3.4028235e+38, initial_weight_distribution='uniform_adaptive', initial_weight_scale=1.0, initial_weights=None, initial_biases=None, loss='automatic', distribution='auto', quantile_alpha=0.5, tweedie_power=1.5, huber_alpha=0.9, score_interval=5.0, score_training_samples=10000, score_validation_samples=0, score_duty_cycle=0.1, classification_stop=0.0, regression_stop=1e-06, stopping_rounds=5, stopping_metric='auto', stopping_tolerance=0.0, max_runtime_secs=0.0, score_validation_sampling='uniform', diagnostics=True, fast_mode=True, force_load_balance=True, variable_importances=True, replicate_training_data=True, single_node_mode=False, shuffle_training_data=False, missing_values_handling='mean_imputation', quiet_mode=False, autoencoder=False, sparse=False, col_major=False, average_activation=0.0, sparsity_beta=0.0, max_categorical_features=2147483647, reproducible=False, export_weights_and_biases=False, mini_batch_size=1, categorical_encoding='auto', elastic_averaging=False, elastic_averaging_moving_rate=0.9, elastic_averaging_regularization=0.001, export_checkpoints_dir=None, auc_type='auto')[source]

Bases: h2o.estimators.estimator_base.H2OEstimator

Deep Learning

Build a Deep Neural Network model using CPUs Builds a feed-forward multilayer artificial neural network on an H2OFrame

Examples

>>> from h2o.estimators.deeplearning import H2ODeepLearningEstimator
>>> rows = [[1,2,3,4,0], [2,1,2,4,1], [2,1,4,2,1],
...         [0,1,2,34,1], [2,3,4,1,0]] * 50
>>> fr = h2o.H2OFrame(rows)
>>> fr[4] = fr[4].asfactor()
>>> model = H2ODeepLearningEstimator()
>>> model.train(x=range(4), y=4, training_frame=fr)
>>> model.logloss()
property activation

Activation function.

Type: Literal["tanh", "tanh_with_dropout", "rectifier", "rectifier_with_dropout", "maxout", "maxout_with_dropout"], defaults to "rectifier".

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "cylinders"
>>> cars_dl = H2ODeepLearningEstimator(activation="tanh")
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cars_dl.mse()
property adaptive_rate

Adaptive learning rate.

Type: bool, defaults to True.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "cylinders"
>>> cars_dl = H2ODeepLearningEstimator(adaptive_rate=True)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cars_dl.mse()
property auc_type

Set default multinomial AUC type.

Type: Literal["auto", "none", "macro_ovr", "weighted_ovr", "macro_ovo", "weighted_ovo"], defaults to "auto".

property autoencoder

Auto-Encoder.

Type: bool, defaults to False.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "cylinders"
>>> cars_dl = H2ODeepLearningEstimator(autoencoder=True)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cars_dl.mse()
property average_activation

Average activation for sparse auto-encoder. #Experimental

Type: float, defaults to 0.0.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "cylinders"
>>> cars_dl = H2ODeepLearningEstimator(average_activation=1.5,
...                                    seed=1234)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cars_dl.mse()
property balance_classes

Balance training data class counts via over/under-sampling (for imbalanced data).

Type: bool, defaults to False.

Examples

>>> covtype = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/covtype/covtype.20k.data")
>>> covtype[54] = covtype[54].asfactor()
>>> predictors = covtype.columns[0:54]
>>> response = 'C55'
>>> train, valid = covtype.split_frame(ratios=[.8], seed=1234)
>>> cov_dl = H2ODeepLearningEstimator(balance_classes=True,
...                                   seed=1234)
>>> cov_dl.train(x=predictors,
...              y=response,
...              training_frame=train,
...              validation_frame=valid)
>>> cov_dl.mse()
property categorical_encoding

Encoding scheme for categorical features

Type: Literal["auto", "enum", "one_hot_internal", "one_hot_explicit", "binary", "eigen", "label_encoder", "sort_by_response", "enum_limited"], defaults to "auto".

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Year"]= airlines["Year"].asfactor()
>>> airlines["Month"]= airlines["Month"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> train, valid= airlines.split_frame(ratios=[.8], seed=1234)
>>> encoding = "one_hot_internal"
>>> airlines_dl = H2ODeepLearningEstimator(categorical_encoding=encoding,
...                                        seed=1234)
>>> airlines_dl.train(x=predictors,
...                   y=response,
...                   training_frame=train,
...                   validation_frame=valid)
>>> airlines_dl.mse()
property checkpoint

Model checkpoint to resume training with.

Type: Union[None, str, H2OEstimator].

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_dl = H2ODeepLearningEstimator(activation="tanh",
...                                    autoencoder=True,
...                                    seed=1234,
...                                    model_id="cars_dl")
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cars_dl.mse()
>>> cars_cont = H2ODeepLearningEstimator(checkpoint=cars_dl,
...                                      seed=1234)
>>> cars_cont.train(x=predictors,
...                 y=response,
...                 training_frame=train,
...                 validation_frame=valid)
>>> cars_cont.mse()
property class_sampling_factors

Desired over/under-sampling ratios per class (in lexicographic order). If not specified, sampling factors will be automatically computed to obtain class balance during training. Requires balance_classes.

Type: List[float].

Examples

>>> covtype = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/covtype/covtype.20k.data")
>>> covtype[54] = covtype[54].asfactor()
>>> predictors = covtype.columns[0:54]
>>> response = 'C55'
>>> train, valid = covtype.split_frame(ratios=[.8], seed=1234)
>>> sample_factors = [1., 0.5, 1., 1., 1., 1., 1.]
>>> cars_dl = H2ODeepLearningEstimator(balance_classes=True,
...                                    class_sampling_factors=sample_factors,
...                                    seed=1234)
>>> cov_dl.train(x=predictors,
...              y=response,
...              training_frame=train,
...              validation_frame=valid)
>>> cov_dl.mse()
property classification_stop

Stopping criterion for classification error fraction on training data (-1 to disable).

Type: float, defaults to 0.0.

Examples

>>> covtype = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/covtype/covtype.20k.data")
>>> covtype[54] = covtype[54].asfactor()
>>> predictors = covtype.columns[0:54]
>>> response = 'C55'
>>> train, valid = covtype.split_frame(ratios=[.8], seed=1234)
>>> cars_dl = H2ODeepLearningEstimator(classification_stop=1.5,
...                                    seed=1234)
>>> cov_dl.train(x=predictors,
...              y=response,
...              training_frame=train,
...              validation_frame=valid)
>>> cov_dl.mse()
property col_major

#DEPRECATED Use a column major weight matrix for input layer. Can speed up forward propagation, but might slow down backpropagation.

Type: bool, defaults to False.

property diagnostics

Enable diagnostics for hidden layers.

Type: bool, defaults to True.

Examples

>>> covtype = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/covtype/covtype.20k.data")
>>> covtype[54] = covtype[54].asfactor()
>>> predictors = covtype.columns[0:54]
>>> response = 'C55'
>>> train, valid = covtype.split_frame(ratios=[.8], seed=1234)
>>> cars_dl = H2ODeepLearningEstimator(diagnostics=True,
...                                    seed=1234)  
>>> cov_dl.train(x=predictors,
...              y=response,
...              training_frame=train,
...              validation_frame=valid)
>>> cov_dl.mse()
property distribution

Distribution function

Type: Literal["auto", "bernoulli", "multinomial", "gaussian", "poisson", "gamma", "tweedie", "laplace", "quantile", "huber"], defaults to "auto".

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "cylinders"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_dl = H2ODeepLearningEstimator(distribution="poisson",
...                                    seed=1234)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cars_dl.mse()
property elastic_averaging

Elastic averaging between compute nodes can improve distributed model convergence. #Experimental

Type: bool, defaults to False.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "cylinders"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_dl = H2ODeepLearningEstimator(elastic_averaging=True,
...                                    seed=1234)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cars_dl.mse()
property elastic_averaging_moving_rate

Elastic averaging moving rate (only if elastic averaging is enabled).

Type: float, defaults to 0.9.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "cylinders"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_dl = H2ODeepLearningEstimator(elastic_averaging_moving_rate=.8,
...                                    seed=1234)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cars_dl.mse()
property elastic_averaging_regularization

Elastic averaging regularization strength (only if elastic averaging is enabled).

Type: float, defaults to 0.001.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "cylinders"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_dl = H2ODeepLearningEstimator(elastic_averaging_regularization=.008,
...                                    seed=1234)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cars_dl.mse()
property epochs

How many times the dataset should be iterated (streamed), can be fractional.

Type: float, defaults to 10.0.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "cylinders"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_dl = H2ODeepLearningEstimator(epochs=15,
...                                    seed=1234)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cars_dl.mse()
property epsilon

Adaptive learning rate smoothing factor (to avoid divisions by zero and allow progress).

Type: float, defaults to 1e-08.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "cylinders"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_dl = H2ODeepLearningEstimator(epsilon=1e-6,
...                                    seed=1234)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cars_dl.mse()
property export_checkpoints_dir

Automatically export generated models to this directory.

Type: str.

Examples

>>> import tempfile
>>> from os import listdir
>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "cylinders"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> checkpoints_dir = tempfile.mkdtemp()
>>> cars_dl = H2ODeepLearningEstimator(export_checkpoints_dir=checkpoints_dir,
...                                    seed=1234)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> len(listdir(checkpoints_dir))
property export_weights_and_biases

Whether to export Neural Network weights and biases to H2O Frames.

Type: bool, defaults to False.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "cylinders"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_dl = H2ODeepLearningEstimator(export_weights_and_biases=True,
...                                    seed=1234)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cars_dl.mse()
property fast_mode

Enable fast mode (minor approximation in back-propagation).

Type: bool, defaults to True.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "cylinders"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_dl = H2ODeepLearningEstimator(fast_mode=False,
...                                    seed=1234)          
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cars_dl.mse()
property fold_assignment

Cross-validation fold assignment scheme, if fold_column is not specified. The ‘Stratified’ option will stratify the folds based on the response variable, for classification problems.

Type: Literal["auto", "random", "modulo", "stratified"], defaults to "auto".

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "cylinders"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_dl = H2ODeepLearningEstimator(fold_assignment="Random",
...                                    nfolds=5,
...                                    seed=1234)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cars_dl.mse()
property fold_column

Column with cross-validation fold index assignment per observation.

Type: str.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "cylinders"
>>> fold_numbers = cars.kfold_column(n_folds=5, seed=1234)
>>> fold_numbers.set_names(["fold_numbers"])
>>> cars = cars.cbind(fold_numbers)
>>> print(cars['fold_numbers'])
>>> cars_dl = H2ODeepLearningEstimator(seed=1234)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=cars,
...               fold_column="fold_numbers")
>>> cars_dl.mse()
property force_load_balance

Force extra load balancing to increase training speed for small datasets (to keep all cores busy).

Type: bool, defaults to True.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "cylinders"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_dl = H2ODeepLearningEstimator(force_load_balance=False,
...                                    seed=1234)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cars_dl.mse()
property hidden

Hidden layer sizes (e.g. [100, 100]).

Type: List[int], defaults to [200, 200].

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "cylinders"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_dl = H2ODeepLearningEstimator(hidden=[100,100],
...                                    seed=1234) 
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cars_dl.mse()
property hidden_dropout_ratios

Hidden layer dropout ratios (can improve generalization), specify one value per hidden layer, defaults to 0.5.

Type: List[float].

Examples

>>> train = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/bigdata/laptop/mnist/train.csv.gz")
>>> valid = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/bigdata/laptop/mnist/test.csv.gz")
>>> features = list(range(0,784))
>>> target = 784
>>> train[target] = train[target].asfactor()
>>> valid[target] = valid[target].asfactor()
>>> model = H2ODeepLearningEstimator(epochs=20,
...                                  hidden=[200,200],
...                                  hidden_dropout_ratios=[0.5,0.5],
...                                  seed=1234,
...                                  activation='tanhwithdropout')
>>> model.train(x=features,
...             y=target,
...             training_frame=train,
...             validation_frame=valid)
>>> model.mse()
property huber_alpha

Desired quantile for Huber/M-regression (threshold between quadratic and linear loss, must be between 0 and 1).

Type: float, defaults to 0.9.

Examples

>>> insurance = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/glm_test/insurance.csv")
>>> predictors = insurance.columns[0:4]
>>> response = 'Claims'
>>> insurance['Group'] = insurance['Group'].asfactor()
>>> insurance['Age'] = insurance['Age'].asfactor()
>>> train, valid = insurance.split_frame(ratios=[.8], seed=1234)
>>> insurance_dl = H2ODeepLearningEstimator(distribution="huber",
...                                         huber_alpha=0.9,
...                                         seed=1234)
>>> insurance_dl.train(x=predictors,
...                    y=response,
...                    training_frame=train,
...                    validation_frame=valid)
>>> insurance_dl.mse()
property ignore_const_cols

Ignore constant columns.

Type: bool, defaults to True.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> cars["const_1"] = 6
>>> cars["const_2"] = 7
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_dl = H2ODeepLearningEstimator(seed=1234,
...                                    ignore_const_cols=True)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cars_dl.auc()
property ignored_columns

Names of columns to ignore for training.

Type: List[str].

property initial_biases

A list of H2OFrame ids to initialize the bias vectors of this model with.

Type: List[Union[None, str, H2OFrame]].

Examples

>>> iris = h2o.import_file("http://h2o-public-test-data.s3.amazonaws.com/smalldata/iris/iris.csv")
>>> dl1 = H2ODeepLearningEstimator(hidden=[10,10],
...                                export_weights_and_biases=True)
>>> dl1.train(x=list(range(4)), y=4, training_frame=iris)
>>> p1 = dl1.model_performance(iris).logloss()
>>> ll1 = dl1.predict(iris)
>>> print(p1)
>>> w1 = dl1.weights(0)
>>> w2 = dl1.weights(1)
>>> w3 = dl1.weights(2)
>>> b1 = dl1.biases(0)
>>> b2 = dl1.biases(1)
>>> b3 = dl1.biases(2)
>>> dl2 = H2ODeepLearningEstimator(hidden=[10,10],
...                                initial_weights=[w1, w2, w3],
...                                initial_biases=[b1, b2, b3],
...                                epochs=0)
>>> dl2.train(x=list(range(4)), y=4, training_frame=iris)
>>> dl2.initial_biases
property initial_weight_distribution

Initial weight distribution.

Type: Literal["uniform_adaptive", "uniform", "normal"], defaults to "uniform_adaptive".

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_dl = H2ODeepLearningEstimator(initial_weight_distribution="Uniform",
...                                    seed=1234)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cars_dl.auc()
property initial_weight_scale

Uniform: -value…value, Normal: stddev.

Type: float, defaults to 1.0.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_dl = H2ODeepLearningEstimator(initial_weight_scale=1.5,
...                                    seed=1234) 
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cars_dl.auc()
property initial_weights

A list of H2OFrame ids to initialize the weight matrices of this model with.

Type: List[Union[None, str, H2OFrame]].

Examples

>>> iris = h2o.import_file("http://h2o-public-test-data.s3.amazonaws.com/smalldata/iris/iris.csv")
>>> dl1 = H2ODeepLearningEstimator(hidden=[10,10],
...                                export_weights_and_biases=True)
>>> dl1.train(x=list(range(4)), y=4, training_frame=iris)
>>> p1 = dl1.model_performance(iris).logloss()
>>> ll1 = dl1.predict(iris)
>>> print(p1)
>>> w1 = dl1.weights(0)
>>> w2 = dl1.weights(1)
>>> w3 = dl1.weights(2)
>>> b1 = dl1.biases(0)
>>> b2 = dl1.biases(1)
>>> b3 = dl1.biases(2)
>>> dl2 = H2ODeepLearningEstimator(hidden=[10,10],
...                                initial_weights=[w1, w2, w3],
...                                initial_biases=[b1, b2, b3],
...                                epochs=0)
>>> dl2.train(x=list(range(4)), y=4, training_frame=iris)
>>> dl2.initial_weights
property input_dropout_ratio

Input layer dropout ratio (can improve generalization, try 0.1 or 0.2).

Type: float, defaults to 0.0.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_dl = H2ODeepLearningEstimator(input_dropout_ratio=0.2,
...                                    seed=1234) 
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cars_dl.auc()
property keep_cross_validation_fold_assignment

Whether to keep the cross-validation fold assignment.

Type: bool, defaults to False.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> cars_dl = H2ODeepLearningEstimator(keep_cross_validation_fold_assignment=True,
...                                    nfolds=5,
...                                    seed=1234)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=cars)
>>> print(cars_dl.cross_validation_fold_assignment())
property keep_cross_validation_models

Whether to keep the cross-validation models.

Type: bool, defaults to True.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> cars_dl = H2ODeepLearningEstimator(keep_cross_validation_models=True,
...                                    nfolds=5,
...                                    seed=1234)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=cars)
>>> print(cars_dl.cross_validation_models())
property keep_cross_validation_predictions

Whether to keep the predictions of the cross-validation models.

Type: bool, defaults to False.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> cars_dl = H2ODeepLearningEstimator(keep_cross_validation_predictions=True,
...                                    nfolds=5,
...                                    seed=1234)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=cars)
>>> print(cars_dl.cross_validation_predictions())
property l1

L1 regularization (can add stability and improve generalization, causes many weights to become 0).

Type: float, defaults to 0.0.

Examples

>>> covtype = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/covtype/covtype.20k.data")
>>> covtype[54] = covtype[54].asfactor()
>>> hh_imbalanced = H2ODeepLearningEstimator(l1=1e-5,
...                                          activation="Rectifier",
...                                          loss="CrossEntropy",
...                                          hidden=[200,200],
...                                          epochs=1,
...                                          balance_classes=False,
...                                          reproducible=True,
...                                          seed=1234)
>>> hh_imbalanced.train(x=list(range(54)),y=54, training_frame=covtype)
>>> hh_imbalanced.mse()
property l2

L2 regularization (can add stability and improve generalization, causes many weights to be small.

Type: float, defaults to 0.0.

Examples

>>> covtype = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/covtype/covtype.20k.data")
>>> covtype[54] = covtype[54].asfactor()
>>> hh_imbalanced = H2ODeepLearningEstimator(l2=1e-5,
...                                          activation="Rectifier",
...                                          loss="CrossEntropy",
...                                          hidden=[200,200],
...                                          epochs=1,
...                                          balance_classes=False,
...                                          reproducible=True,
...                                          seed=1234)
>>> hh_imbalanced.train(x=list(range(54)),y=54, training_frame=covtype)
>>> hh_imbalanced.mse()
property loss

Loss function.

Type: Literal["automatic", "cross_entropy", "quadratic", "huber", "absolute", "quantile"], defaults to "automatic".

Examples

>>> covtype = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/covtype/covtype.20k.data")
>>> covtype[54] = covtype[54].asfactor()
>>> hh_imbalanced = H2ODeepLearningEstimator(l1=1e-5,
...                                          activation="Rectifier",
...                                          loss="CrossEntropy",
...                                          hidden=[200,200],
...                                          epochs=1,
...                                          balance_classes=False,
...                                          reproducible=True,
...                                          seed=1234)
>>> hh_imbalanced.train(x=list(range(54)),y=54, training_frame=covtype)
>>> hh_imbalanced.mse()
property max_after_balance_size

Maximum relative size of the training data after balancing class counts (can be less than 1.0). Requires balance_classes.

Type: float, defaults to 5.0.

Examples

>>> covtype = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/covtype/covtype.20k.data")
>>> covtype[54] = covtype[54].asfactor()
>>> predictors = covtype.columns[0:54]
>>> response = 'C55'
>>> train, valid = covtype.split_frame(ratios=[.8], seed=1234)
>>> max = .85
>>> cov_dl = H2ODeepLearningEstimator(balance_classes=True,
...                                   max_after_balance_size=max,
...                                   seed=1234)
>>> cov_dl.train(x=predictors,
...              y=response,
...              training_frame=train,
...              validation_frame=valid)
>>> cov_dl.logloss()
property max_categorical_features

Max. number of categorical features, enforced via hashing. #Experimental

Type: int, defaults to 2147483647.

Examples

>>> covtype = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/covtype/covtype.20k.data")
>>> covtype[54] = covtype[54].asfactor()
>>> predictors = covtype.columns[0:54]
>>> response = 'C55'
>>> train, valid = covtype.split_frame(ratios=[.8], seed=1234)
>>> cov_dl = H2ODeepLearningEstimator(balance_classes=True,
...                                   max_categorical_features=2147483647,
...                                   seed=1234)
>>> cov_dl.train(x=predictors,
...              y=response,
...              training_frame=train,
...              validation_frame=valid)
>>> cov_dl.logloss()
property max_confusion_matrix_size

[Deprecated] Maximum size (# classes) for confusion matrices to be printed in the Logs.

Type: int, defaults to 20.

property max_runtime_secs

Maximum allowed runtime in seconds for model training. Use 0 to disable.

Type: float, defaults to 0.0.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_dl = H2ODeepLearningEstimator(max_runtime_secs=10,
...                                    seed=1234) 
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cars_dl.auc()
property max_w2

Constraint for squared sum of incoming weights per unit (e.g. for Rectifier).

Type: float, defaults to 3.4028235e+38.

Examples

>>> covtype = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/covtype/covtype.20k.data")
>>> covtype[54] = covtype[54].asfactor()
>>> predictors = covtype.columns[0:54]
>>> response = 'C55'
>>> train, valid = covtype.split_frame(ratios=[.8], seed=1234)
>>> cov_dl = H2ODeepLearningEstimator(activation="RectifierWithDropout",
...                                   hidden=[10,10],
...                                   epochs=10,
...                                   input_dropout_ratio=0.2,
...                                   l1=1e-5,
...                                   max_w2=10.5,
...                                   stopping_rounds=0)
>>> cov_dl.train(x=predictors,
...              y=response,
...              training_frame=train,
...              validation_frame=valid)
>>> cov_dl.mse()
property mini_batch_size

Mini-batch size (smaller leads to better fit, larger can speed up and generalize better).

Type: int, defaults to 1.

Examples

>>> covtype = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/covtype/covtype.20k.data")
>>> covtype[54] = covtype[54].asfactor()
>>> predictors = covtype.columns[0:54]
>>> response = 'C55'
>>> train, valid = covtype.split_frame(ratios=[.8], seed=1234)
>>> cov_dl = H2ODeepLearningEstimator(activation="RectifierWithDropout",
...                                   hidden=[10,10],
...                                   epochs=10,
...                                   input_dropout_ratio=0.2,
...                                   l1=1e-5,
...                                   max_w2=10.5,
...                                   stopping_rounds=0)
...                                   mini_batch_size=35
>>> cov_dl.train(x=predictors,
...              y=response,
...              training_frame=train,
...              validation_frame=valid)
>>> cov_dl.mse()
property missing_values_handling

Handling of missing values. Either MeanImputation or Skip.

Type: Literal["mean_imputation", "skip"], defaults to "mean_imputation".

Examples

>>> boston = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/gbm_test/BostonHousing.csv")
>>> predictors = boston.columns[:-1]
>>> response = "medv"
>>> boston['chas'] = boston['chas'].asfactor()
>>> boston.insert_missing_values()
>>> train, valid = boston.split_frame(ratios=[.8])
>>> boston_dl = H2ODeepLearningEstimator(missing_values_handling="skip")
>>> boston_dl.train(x=predictors,
...                 y=response,
...                 training_frame=train,
...                 validation_frame=valid)
>>> boston_dl.mse()
property momentum_ramp

Number of training samples for which momentum increases.

Type: float, defaults to 1000000.0.

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> predictors = ["Year","Month","DayofMonth","DayOfWeek","CRSDepTime",
...               "CRSArrTime","UniqueCarrier","FlightNum"]
>>> response_col = "IsDepDelayed"
>>> airlines_dl = H2ODeepLearningEstimator(hidden=[200,200],
...                                        activation="Rectifier",
...                                        input_dropout_ratio=0.0,
...                                        momentum_start=0.9,
...                                        momentum_stable=0.99,
...                                        momentum_ramp=1e7,
...                                        epochs=100,
...                                        stopping_rounds=4,
...                                        train_samples_per_iteration=30000,
...                                        mini_batch_size=32,
...                                        score_duty_cycle=0.25,
...                                        score_interval=1)
>>> airlines_dl.train(x=predictors,
...                   y=response_col,
...                   training_frame=airlines)
>>> airlines_dl.mse()
property momentum_stable

Final momentum after the ramp is over (try 0.99).

Type: float, defaults to 0.0.

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> predictors = ["Year","Month","DayofMonth","DayOfWeek","CRSDepTime",
...               "CRSArrTime","UniqueCarrier","FlightNum"]
>>> response_col = "IsDepDelayed"
>>> airlines_dl = H2ODeepLearningEstimator(hidden=[200,200],
...                                        activation="Rectifier",
...                                        input_dropout_ratio=0.0,
...                                        momentum_start=0.9,
...                                        momentum_stable=0.99,
...                                        momentum_ramp=1e7,
...                                        epochs=100,
...                                        stopping_rounds=4,
...                                        train_samples_per_iteration=30000,
...                                        mini_batch_size=32,
...                                        score_duty_cycle=0.25,
...                                        score_interval=1)
>>> airlines_dl.train(x=predictors,
...                   y=response_col,
...                   training_frame=airlines)
>>> airlines_dl.mse()
property momentum_start

Initial momentum at the beginning of training (try 0.5).

Type: float, defaults to 0.0.

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> predictors = ["Year","Month","DayofMonth","DayOfWeek","CRSDepTime",
...               "CRSArrTime","UniqueCarrier","FlightNum"]
>>> response_col = "IsDepDelayed"
>>> airlines_dl = H2ODeepLearningEstimator(hidden=[200,200],
...                                        activation="Rectifier",
...                                        input_dropout_ratio=0.0,
...                                        momentum_start=0.9,
...                                        momentum_stable=0.99,
...                                        momentum_ramp=1e7,
...                                        epochs=100,
...                                        stopping_rounds=4,
...                                        train_samples_per_iteration=30000,
...                                        mini_batch_size=32,
...                                        score_duty_cycle=0.25,
...                                        score_interval=1)
>>> airlines_dl.train(x=predictors,
...                   y=response_col,
...                   training_frame=airlines)
>>> airlines_dl.mse()
property nesterov_accelerated_gradient

Use Nesterov accelerated gradient (recommended).

Type: bool, defaults to True.

Examples

>>> train = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/bigdata/laptop/mnist/train.csv.gz")
>>> test = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/bigdata/laptop/mnist/test.csv.gz")
>>> predictors = list(range(0,784))
>>> resp = 784
>>> train[resp] = train[resp].asfactor()
>>> test[resp] = test[resp].asfactor()
>>> nclasses = train[resp].nlevels()[0]
>>> model = H2ODeepLearningEstimator(activation="RectifierWithDropout",
...                                  adaptive_rate=False,
...                                  rate=0.01,
...                                  rate_decay=0.9,
...                                  rate_annealing=1e-6,
...                                  momentum_start=0.95,
...                                  momentum_ramp=1e5,
...                                  momentum_stable=0.99,
...                                  nesterov_accelerated_gradient=False,
...                                  input_dropout_ratio=0.2,
...                                  train_samples_per_iteration=20000,
...                                  classification_stop=-1,
...                                  l1=1e-5) 
>>> model.train (x=predictors,
...              y=resp,
...              training_frame=train,
...              validation_frame=test)
>>> model.model_performance()
property nfolds

Number of folds for K-fold cross-validation (0 to disable or >= 2).

Type: int, defaults to 0.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> cars_dl = H2ODeepLearningEstimator(nfolds=5, seed=1234)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=cars)
>>> cars_dl.auc()
property offset_column

Offset column. This will be added to the combination of columns before applying the link function.

Type: str.

Examples

>>> boston = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/gbm_test/BostonHousing.csv")
>>> predictors = boston.columns[:-1]
>>> response = "medv"
>>> boston['chas'] = boston['chas'].asfactor()
>>> boston["offset"] = boston["medv"].log()
>>> train, valid = boston.split_frame(ratios=[.8], seed=1234)
>>> boston_dl = H2ODeepLearningEstimator(offset_column="offset",
...                                      seed=1234)
>>> boston_dl.train(x=predictors,
...                 y=response,
...                 training_frame=train,
...                 validation_frame=valid)
>>> boston_dl.mse()
property overwrite_with_best_model

If enabled, override the final model with the best model found during training.

Type: bool, defaults to True.

Examples

>>> boston = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/gbm_test/BostonHousing.csv")
>>> predictors = boston.columns[:-1]
>>> response = "medv"
>>> boston['chas'] = boston['chas'].asfactor()
>>> boston["offset"] = boston["medv"].log()
>>> train, valid = boston.split_frame(ratios=[.8], seed=1234)
>>> boston_dl = H2ODeepLearningEstimator(overwrite_with_best_model=True,
...                                      seed=1234)
>>> boston_dl.train(x=predictors,
...                 y=response,
...                 training_frame=train,
...                 validation_frame=valid)
>>> boston_dl.mse()
property pretrained_autoencoder

Pretrained autoencoder model to initialize this model with.

Type: Union[None, str, H2OEstimator].

Examples

>>> from h2o.estimators.deeplearning import H2OAutoEncoderEstimator
>>> resp = 784
>>> nfeatures = 20
>>> train = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/bigdata/laptop/mnist/train.csv.gz")
>>> test = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/bigdata/laptop/mnist/test.csv.gz")
>>> train[resp] = train[resp].asfactor()
>>> test[resp] = test[resp].asfactor()
>>> sid = train[0].runif(0)
>>> train_unsupervised = train[sid>=0.5]
>>> train_unsupervised.pop(resp)
>>> train_supervised = train[sid<0.5]
>>> ae_model = H2OAutoEncoderEstimator(activation="Tanh",
...                                    hidden=[nfeatures],
...                                    model_id="ae_model",
...                                    epochs=1,
...                                    ignore_const_cols=False,
...                                    reproducible=True,
...                                    seed=1234)
>>> ae_model.train(list(range(resp)), training_frame=train_unsupervised)
>>> ae_model.mse()
>>> pretrained_model = H2ODeepLearningEstimator(activation="Tanh",
...                                             hidden=[nfeatures],
...                                             epochs=1,
...                                             reproducible = True,
...                                             seed=1234,
...                                             ignore_const_cols=False,
...                                             pretrained_autoencoder="ae_model")
>>> pretrained_model.train(list(range(resp)), resp,
...                        training_frame=train_supervised,
...                        validation_frame=test)
>>> pretrained_model.mse()
property quantile_alpha

Desired quantile for Quantile regression, must be between 0 and 1.

Type: float, defaults to 0.5.

Examples

>>> boston = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/gbm_test/BostonHousing.csv")
>>> predictors = boston.columns[:-1]
>>> response = "medv"
>>> boston['chas'] = boston['chas'].asfactor()
>>> train, valid = boston.split_frame(ratios=[.8], seed=1234)
>>> boston_dl = H2ODeepLearningEstimator(distribution="quantile",
...                                      quantile_alpha=.8,
...                                      seed=1234)
>>> boston_dl.train(x=predictors,
...                 y=response,
...                 training_frame=train,
...                 validation_frame=valid)
>>> boston_dl.mse()
property quiet_mode

Enable quiet mode for less output to standard output.

Type: bool, defaults to False.

Examples

>>> titanic = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/gbm_test/titanic.csv")
>>> titanic['survived'] = titanic['survived'].asfactor()
>>> predictors = titanic.columns
>>> del predictors[1:3]
>>> response = 'survived'
>>> train, valid = titanic.split_frame(ratios=[.8], seed=1234)
>>> titanic_dl = H2ODeepLearningEstimator(quiet_mode=True,
...                                       seed=1234)
>>> titanic_dl.train(x=predictors,
...                  y=response,
...                  training_frame=train,
...                  validation_frame=valid)
>>> titanic_dl.mse()
property rate

Learning rate (higher => less stable, lower => slower convergence).

Type: float, defaults to 0.005.

Examples

>>> train = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/bigdata/laptop/mnist/train.csv.gz")
>>> test = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/bigdata/laptop/mnist/test.csv.gz")
>>> predictors = list(range(0,784))
>>> resp = 784
>>> train[resp] = train[resp].asfactor()
>>> test[resp] = test[resp].asfactor()
>>> nclasses = train[resp].nlevels()[0]
>>> model = H2ODeepLearningEstimator(activation="RectifierWithDropout",
...                                  adaptive_rate=False,
...                                  rate=0.01,
...                                  rate_decay=0.9,
...                                  rate_annealing=1e-6,
...                                  momentum_start=0.95,
...                                  momentum_ramp=1e5,
...                                  momentum_stable=0.99,
...                                  nesterov_accelerated_gradient=False,
...                                  input_dropout_ratio=0.2,
...                                  train_samples_per_iteration=20000,
...                                  classification_stop=-1,
...                                  l1=1e-5)
>>> model.train (x=predictors,y=resp, training_frame=train, validation_frame=test)
>>> model.model_performance(valid=True)
property rate_annealing

Learning rate annealing: rate / (1 + rate_annealing * samples).

Type: float, defaults to 1e-06.

Examples

>>> train = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/bigdata/laptop/mnist/train.csv.gz")
>>> test = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/bigdata/laptop/mnist/test.csv.gz")
>>> predictors = list(range(0,784))
>>> resp = 784
>>> train[resp] = train[resp].asfactor()
>>> test[resp] = test[resp].asfactor()
>>> nclasses = train[resp].nlevels()[0]
>>> model = H2ODeepLearningEstimator(activation="RectifierWithDropout",
...                                  adaptive_rate=False,
...                                  rate=0.01,
...                                  rate_decay=0.9,
...                                  rate_annealing=1e-6,
...                                  momentum_start=0.95,
...                                  momentum_ramp=1e5,
...                                  momentum_stable=0.99,
...                                  nesterov_accelerated_gradient=False,
...                                  input_dropout_ratio=0.2,
...                                  train_samples_per_iteration=20000,
...                                  classification_stop=-1,
...                                  l1=1e-5)
>>> model.train (x=predictors,
...              y=resp,
...              training_frame=train,
...              validation_frame=test)
>>> model.mse()
property rate_decay

Learning rate decay factor between layers (N-th layer: rate * rate_decay ^ (n - 1).

Type: float, defaults to 1.0.

Examples

>>> train = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/bigdata/laptop/mnist/train.csv.gz")
>>> test = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/bigdata/laptop/mnist/test.csv.gz")
>>> predictors = list(range(0,784))
>>> resp = 784
>>> train[resp] = train[resp].asfactor()
>>> test[resp] = test[resp].asfactor()
>>> nclasses = train[resp].nlevels()[0]
>>> model = H2ODeepLearningEstimator(activation="RectifierWithDropout",
...                                  adaptive_rate=False,
...                                  rate=0.01,
...                                  rate_decay=0.9,
...                                  rate_annealing=1e-6,
...                                  momentum_start=0.95,
...                                  momentum_ramp=1e5,
...                                  momentum_stable=0.99,
...                                  nesterov_accelerated_gradient=False,
...                                  input_dropout_ratio=0.2,
...                                  train_samples_per_iteration=20000,
...                                  classification_stop=-1,
...                                  l1=1e-5)
>>> model.train (x=predictors,
...              y=resp,
...              training_frame=train,
...              validation_frame=test)
>>> model.model_performance()
property regression_stop

Stopping criterion for regression error (MSE) on training data (-1 to disable).

Type: float, defaults to 1e-06.

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Year"]= airlines["Year"].asfactor()
>>> airlines["Month"]= airlines["Month"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> train, valid= airlines.split_frame(ratios=[.8], seed=1234)
>>> airlines_dl = H2ODeepLearningEstimator(regression_stop=1e-6,
...                                        seed=1234)
>>> airlines_dl.train(x=predictors,
...                   y=response,
...                   training_frame=train,
...                   validation_frame=valid)
>>> airlines_dl.auc()
property replicate_training_data

Replicate the entire training dataset onto every node for faster training on small datasets.

Type: bool, defaults to True.

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Year"]= airlines["Year"].asfactor()
>>> airlines["Month"]= airlines["Month"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> airlines_dl = H2ODeepLearningEstimator(replicate_training_data=False)
>>> airlines_dl.train(x=predictors,
...                   y=response,
...                   training_frame=airlines) 
>>> airlines_dl.auc()
property reproducible

Force reproducibility on small data (will be slow - only uses 1 thread).

Type: bool, defaults to False.

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Year"]= airlines["Year"].asfactor()
>>> airlines["Month"]= airlines["Month"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> train, valid= airlines.split_frame(ratios=[.8], seed=1234)
>>> airlines_dl = H2ODeepLearningEstimator(reproducible=True)
>>> airlines_dl.train(x=predictors,
...                   y=response,
...                   training_frame=train,
...                   validation_frame=valid)
>>> airlines_dl.auc()
property response_column

Response variable column.

Type: str.

property rho

Adaptive learning rate time decay factor (similarity to prior updates).

Type: float, defaults to 0.99.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> cars_dl = H2ODeepLearningEstimator(rho=0.9,
...                                    seed=1234)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=cars)
>>> cars_dl.auc()
property score_duty_cycle

Maximum duty cycle fraction for scoring (lower: more training, higher: more scoring).

Type: float, defaults to 0.1.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> cars_dl = H2ODeepLearningEstimator(score_duty_cycle=0.2,
...                                    seed=1234)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=cars)
>>> cars_dl.auc()
property score_each_iteration

Whether to score during each iteration of model training.

Type: bool, defaults to False.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> cars_dl = H2ODeepLearningEstimator(score_each_iteration=True,
...                                    seed=1234) 
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=cars)
>>> cars_dl.auc()
property score_interval

Shortest time interval (in seconds) between model scoring.

Type: float, defaults to 5.0.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> cars_dl = H2ODeepLearningEstimator(score_interval=3,
...                                    seed=1234) 
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=cars)
>>> cars_dl.auc()
property score_training_samples

Number of training set samples for scoring (0 for all).

Type: int, defaults to 10000.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> cars_dl = H2ODeepLearningEstimator(score_training_samples=10000,
...                                    seed=1234)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=cars)
>>> cars_dl.auc()
property score_validation_samples

Number of validation set samples for scoring (0 for all).

Type: int, defaults to 0.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_dl = H2ODeepLearningEstimator(score_validation_samples=3,
...                                    seed=1234)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cars_dl.auc()
property score_validation_sampling

Method used to sample validation dataset for scoring.

Type: Literal["uniform", "stratified"], defaults to "uniform".

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_dl = H2ODeepLearningEstimator(score_validation_sampling="uniform",
...                                    seed=1234)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cars_dl.auc()
property seed

Seed for random numbers (affects sampling) - Note: only reproducible when running single threaded.

Type: int, defaults to -1.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_dl = H2ODeepLearningEstimator(seed=1234)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cars_dl.auc()
property shuffle_training_data

Enable shuffling of training data (recommended if training data is replicated and train_samples_per_iteration is close to #nodes x #rows, of if using balance_classes).

Type: bool, defaults to False.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_dl = H2ODeepLearningEstimator(shuffle_training_data=True,
...                                    seed=1234)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=cars)
>>> cars_dl.auc()
property single_node_mode

Run on a single node for fine-tuning of model parameters.

Type: bool, defaults to False.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_dl = H2ODeepLearningEstimator(single_node_mode=True,
...                                    seed=1234) 
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=cars)
>>> cars_dl.auc()
property sparse

Sparse data handling (more efficient for data with lots of 0 values).

Type: bool, defaults to False.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_dl = H2ODeepLearningEstimator(sparse=True,
...                                    seed=1234)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=cars)
>>> cars_dl.auc()
property sparsity_beta

Sparsity regularization. #Experimental

Type: float, defaults to 0.0.

Examples

>>> from h2o.estimators import H2OAutoEncoderEstimator
>>> resp = 784
>>> nfeatures = 20
>>> train = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/bigdata/laptop/mnist/train.csv.gz")
>>> test = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/bigdata/laptop/mnist/test.csv.gz")
>>> train[resp] = train[resp].asfactor()
>>> test[resp] = test[resp].asfactor()
>>> sid = train[0].runif(0)
>>> train_unsupervised = train[sid>=0.5]
>>> train_unsupervised.pop(resp)
>>> ae_model = H2OAutoEncoderEstimator(activation="Tanh",
...                                    hidden=[nfeatures],
...                                    epochs=1,
...                                    ignore_const_cols=False,
...                                    reproducible=True,
...                                    sparsity_beta=0.5,
...                                    seed=1234)
>>> ae_model.train(list(range(resp)),
...                training_frame=train_unsupervised)
>>> ae_model.mse()
property standardize

If enabled, automatically standardize the data. If disabled, the user must provide properly scaled input data.

Type: bool, defaults to True.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> cars_dl = H2ODeepLearningEstimator(standardize=True,
...                                    seed=1234)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=cars)
>>> cars_dl.auc()
property stopping_metric

Metric to use for early stopping (AUTO: logloss for classification, deviance for regression and anomaly_score for Isolation Forest). Note that custom and custom_increasing can only be used in GBM and DRF with the Python client.

Type: Literal["auto", "deviance", "logloss", "mse", "rmse", "mae", "rmsle", "auc", "aucpr", "lift_top_group", "misclassification", "mean_per_class_error", "custom", "custom_increasing"], defaults to "auto".

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Year"]= airlines["Year"].asfactor()
>>> airlines["Month"]= airlines["Month"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> train, valid= airlines.split_frame(ratios=[.8], seed=1234)
>>> airlines_dl = H2ODeepLearningEstimator(stopping_metric="auc",
...                                        stopping_rounds=3,
...                                        stopping_tolerance=1e-2,
...                                        seed=1234)
>>> airlines_dl.train(x=predictors,
...                   y=response,
...                   training_frame=train,
...                   validation_frame=valid)
>>> airlines_dl.auc()
property stopping_rounds

Early stopping based on convergence of stopping_metric. Stop if simple moving average of length k of the stopping_metric does not improve for k:=stopping_rounds scoring events (0 to disable)

Type: int, defaults to 5.

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Year"]= airlines["Year"].asfactor()
>>> airlines["Month"]= airlines["Month"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> train, valid= airlines.split_frame(ratios=[.8], seed=1234)
>>> airlines_dl = H2ODeepLearningEstimator(stopping_metric="auc",
...                                        stopping_rounds=3,
...                                        stopping_tolerance=1e-2,
...                                        seed=1234)
>>> airlines_dl.train(x=predictors,
...                   y=response,
...                   training_frame=train,
...                   validation_frame=valid)
>>> airlines_dl.auc()
property stopping_tolerance

Relative tolerance for metric-based stopping criterion (stop if relative improvement is not at least this much)

Type: float, defaults to 0.0.

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Year"]= airlines["Year"].asfactor()
>>> airlines["Month"]= airlines["Month"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> train, valid= airlines.split_frame(ratios=[.8], seed=1234)
>>> airlines_dl = H2ODeepLearningEstimator(stopping_metric="auc",
...                                        stopping_rounds=3,
...                                        stopping_tolerance=1e-2,
...                                        seed=1234)
>>> airlines_dl.train(x=predictors,
...                   y=response,
...                   training_frame=train,
...                   validation_frame=valid)
>>> airlines_dl.auc()
property target_ratio_comm_to_comp

Target ratio of communication overhead to computation. Only for multi-node operation and train_samples_per_iteration = -2 (auto-tuning).

Type: float, defaults to 0.05.

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Year"]= airlines["Year"].asfactor()
>>> airlines["Month"]= airlines["Month"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> train, valid= airlines.split_frame(ratios=[.8], seed=1234)
>>> airlines_dl = H2ODeepLearningEstimator(target_ratio_comm_to_comp=0.05,
...                                        seed=1234)
>>> airlines_dl.train(x=predictors,
...                   y=response,
...                   training_frame=train,
...                   validation_frame=valid)
>>> airlines_dl.auc()
property train_samples_per_iteration

Number of training samples (globally) per MapReduce iteration. Special values are 0: one epoch, -1: all available data (e.g., replicated training data), -2: automatic.

Type: int, defaults to -2.

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Year"]= airlines["Year"].asfactor()
>>> airlines["Month"]= airlines["Month"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> train, valid= airlines.split_frame(ratios=[.8], seed=1234)
>>> airlines_dl = H2ODeepLearningEstimator(train_samples_per_iteration=-1,
...                                        epochs=1,
...                                        seed=1234)
>>> airlines_dl.train(x=predictors,
...                   y=response,
...                   training_frame=train,
...                   validation_frame=valid)
>>> airlines_dl.auc()
property training_frame

Id of the training data frame.

Type: Union[None, str, H2OFrame].

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Year"]= airlines["Year"].asfactor()
>>> airlines["Month"]= airlines["Month"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> train, valid= airlines.split_frame(ratios=[.8], seed=1234)
>>> airlines_dl = H2ODeepLearningEstimator()
>>> airlines_dl.train(x=predictors,
...                   y=response,
...                   training_frame=train,
...                   validation_frame=valid)
>>> airlines_dl.auc()
property tweedie_power

Tweedie power for Tweedie regression, must be between 1 and 2.

Type: float, defaults to 1.5.

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Year"]= airlines["Year"].asfactor()
>>> airlines["Month"]= airlines["Month"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> train, valid= airlines.split_frame(ratios=[.8], seed=1234)
>>> airlines_dl = H2ODeepLearningEstimator(tweedie_power=1.5,
...                                        seed=1234) 
>>> airlines_dl.train(x=predictors,
...                   y=response,
...                   training_frame=train,
...                   validation_frame=valid)
>>> airlines_dl.auc()
property use_all_factor_levels

Use all factor levels of categorical variables. Otherwise, the first factor level is omitted (without loss of accuracy). Useful for variable importances and auto-enabled for autoencoder.

Type: bool, defaults to True.

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Year"]= airlines["Year"].asfactor()
>>> airlines["Month"]= airlines["Month"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> train, valid= airlines.split_frame(ratios=[.8], seed=1234)
>>> airlines_dl = H2ODeepLearningEstimator(use_all_factor_levels=True,
...                                        seed=1234)
>>> airlines_dl.train(x=predictors,
...                   y=response,
...                   training_frame=train,
...                   validation_frame=valid)
>>> airlines_dl.mse()
property validation_frame

Id of the validation data frame.

Type: Union[None, str, H2OFrame].

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_dl = H2ODeepLearningEstimator(standardize=True,
...                                    seed=1234)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cars_dl.auc()
property variable_importances

Compute variable importances for input features (Gedeon method) - can be slow for large networks.

Type: bool, defaults to True.

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Year"]= airlines["Year"].asfactor()
>>> airlines["Month"]= airlines["Month"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> train, valid= airlines.split_frame(ratios=[.8], seed=1234)
>>> airlines_dl = H2ODeepLearningEstimator(variable_importances=True,
...                                        seed=1234)
>>> airlines_dl.train(x=predictors,
...                   y=response,
...                   training_frame=train,
...                   validation_frame=valid)
>>> airlines_dl.mse()
property weights_column

Column with observation weights. Giving some observation a weight of zero is equivalent to excluding it from the dataset; giving an observation a relative weight of 2 is equivalent to repeating that row twice. Negative weights are not allowed. Note: Weights are per-row observation weights and do not increase the size of the data frame. This is typically the number of times a row is repeated, but non-integer values are supported as well. During training, rows with higher weights matter more, due to the larger loss function pre-factor. If you set weight = 0 for a row, the returned prediction frame at that row is zero and this is incorrect. To get an accurate prediction, remove all rows with weight == 0.

Type: str.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_dl = H2ODeepLearningEstimator(seed=1234)
>>> cars_dl.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cars_dl.auc()

H2OGeneralizedAdditiveEstimator

class h2o.estimators.gam.H2OGeneralizedAdditiveEstimator(model_id=None, training_frame=None, validation_frame=None, nfolds=0, seed=-1, keep_cross_validation_models=True, keep_cross_validation_predictions=False, keep_cross_validation_fold_assignment=False, fold_assignment='auto', fold_column=None, response_column=None, ignored_columns=None, ignore_const_cols=True, score_each_iteration=False, offset_column=None, weights_column=None, family='auto', tweedie_variance_power=0.0, tweedie_link_power=0.0, theta=0.0, solver='auto', alpha=None, lambda_=None, lambda_search=False, early_stopping=True, nlambdas=-1, standardize=False, missing_values_handling='mean_imputation', plug_values=None, compute_p_values=False, remove_collinear_columns=False, splines_non_negative=None, intercept=True, non_negative=False, max_iterations=-1, objective_epsilon=-1.0, beta_epsilon=0.0001, gradient_epsilon=-1.0, link='family_default', startval=None, prior=-1.0, cold_start=False, lambda_min_ratio=-1.0, beta_constraints=None, max_active_predictors=-1, interactions=None, interaction_pairs=None, obj_reg=-1.0, export_checkpoints_dir=None, stopping_rounds=0, stopping_metric='auto', stopping_tolerance=0.001, balance_classes=False, class_sampling_factors=None, max_after_balance_size=5.0, max_confusion_matrix_size=20, max_runtime_secs=0.0, custom_metric_func=None, num_knots=None, spline_orders=None, knot_ids=None, gam_columns=None, standardize_tp_gam_cols=False, scale_tp_penalty_mat=False, bs=None, scale=None, keep_gam_cols=False, auc_type='auto')[source]

Bases: h2o.estimators.estimator_base.H2OEstimator

Generalized Additive Model

Fits a generalized additive model, specified by a response variable, a set of predictors, and a description of the error distribution.

A subclass of ModelBase is returned. The specific subclass depends on the machine learning task at hand (if it’s binomial classification, then an H2OBinomialModel is returned, if it’s regression then a H2ORegressionModel is returned). The default print-out of the models is shown, but further GAM-specific information can be queried out of the object. Upon completion of the GAM, the resulting object has coefficients, normalized coefficients, residual/null deviance, aic, and a host of model metrics including MSE, AUC (for logistic regression), degrees of freedom, and confusion matrices.

property Lambda

[Deprecated] Use lambda_ instead

property alpha

Distribution of regularization between the L1 (Lasso) and L2 (Ridge) penalties. A value of 1 for alpha represents Lasso regression, a value of 0 produces Ridge regression, and anything in between specifies the amount of mixing between the two. Default value of alpha is 0 when SOLVER = ‘L-BFGS’; 0.5 otherwise.

Type: List[float].

property auc_type

Set default multinomial AUC type.

Type: Literal["auto", "none", "macro_ovr", "weighted_ovr", "macro_ovo", "weighted_ovo"], defaults to "auto".

property balance_classes

Balance training data class counts via over/under-sampling (for imbalanced data).

Type: bool, defaults to False.

property beta_constraints

Beta constraints

Type: Union[None, str, H2OFrame].

property beta_epsilon

Converge if beta changes less (using L-infinity norm) than beta esilon, ONLY applies to IRLSM solver

Type: float, defaults to 0.0001.

property bs

Basis function type for each gam predictors, 0 for cr, 1 for thin plate regression with knots, 2 for monotone I-splines, 3 for NBSplineTypeI M-splines (refer to doc here: https://h2oai.atlassian.net/browse/PUBDEV-8835). If specified, must be the same size as gam_columns

Type: List[int].

property class_sampling_factors

Desired over/under-sampling ratios per class (in lexicographic order). If not specified, sampling factors will be automatically computed to obtain class balance during training. Requires balance_classes.

Type: List[float].

property cold_start

Only applicable to multiple alpha/lambda values when calling GLM from GAM. If false, build the next model for next set of alpha/lambda values starting from the values provided by current model. If true will start GLM model from scratch.

Type: bool, defaults to False.

property compute_p_values

Request p-values computation, p-values work only with IRLSM solver and no regularization

Type: bool, defaults to False.

property custom_metric_func

Reference to custom evaluation function, format: language:keyName=funcName

Type: str.

property early_stopping

Stop early when there is no more relative improvement on train or validation (if provided)

Type: bool, defaults to True.

property export_checkpoints_dir

Automatically export generated models to this directory.

Type: str.

property family

Family. Use binomial for classification with logistic regression, others are for regression problems.

Type: Literal["auto", "gaussian", "binomial", "quasibinomial", "ordinal", "multinomial", "poisson", "gamma", "tweedie", "negativebinomial", "fractionalbinomial"], defaults to "auto".

property fold_assignment

Cross-validation fold assignment scheme, if fold_column is not specified. The ‘Stratified’ option will stratify the folds based on the response variable, for classification problems.

Type: Literal["auto", "random", "modulo", "stratified"], defaults to "auto".

property fold_column

Column with cross-validation fold index assignment per observation.

Type: str.

property gam_columns

Arrays of predictor column names for gam for smoothers using single or multiple predictors like {{‘c1’},{‘c2’,’c3’},{‘c4’},…}

Type: List[List[str]].

property gradient_epsilon

Converge if objective changes less (using L-infinity norm) than this, ONLY applies to L-BFGS solver. Default indicates: If lambda_search is set to False and lambda is equal to zero, the default value of gradient_epsilon is equal to .000001, otherwise the default value is .0001. If lambda_search is set to True, the conditional values above are 1E-8 and 1E-6 respectively.

Type: float, defaults to -1.0.

property ignore_const_cols

Ignore constant columns.

Type: bool, defaults to True.

property ignored_columns

Names of columns to ignore for training.

Type: List[str].

property interaction_pairs

A list of pairwise (first order) column interactions.

Type: List[tuple].

property interactions

A list of predictor column indices to interact. All pairwise combinations will be computed for the list.

Type: List[str].

property intercept

Include constant term in the model

Type: bool, defaults to True.

property keep_cross_validation_fold_assignment

Whether to keep the cross-validation fold assignment.

Type: bool, defaults to False.

property keep_cross_validation_models

Whether to keep the cross-validation models.

Type: bool, defaults to True.

property keep_cross_validation_predictions

Whether to keep the predictions of the cross-validation models.

Type: bool, defaults to False.

property keep_gam_cols

Save keys of model matrix

Type: bool, defaults to False.

property knot_ids

Array storing frame keys of knots. One for each gam column set specified in gam_columns

Type: List[str].

property lambda_

Regularization strength

Type: List[float].

property lambda_min_ratio

Minimum lambda used in lambda search, specified as a ratio of lambda_max (the smallest lambda that drives all coefficients to zero). Default indicates: if the number of observations is greater than the number of variables, then lambda_min_ratio is set to 0.0001; if the number of observations is less than the number of variables, then lambda_min_ratio is set to 0.01.

Type: float, defaults to -1.0.

Use lambda search starting at lambda max, given lambda is then interpreted as lambda min

Type: bool, defaults to False.

Link function.

Type: Literal["family_default", "identity", "logit", "log", "inverse", "tweedie", "ologit"], defaults to "family_default".

property max_active_predictors

Maximum number of active predictors during computation. Use as a stopping criterion to prevent expensive model building with many predictors. Default indicates: If the IRLSM solver is used, the value of max_active_predictors is set to 5000 otherwise it is set to 100000000.

Type: int, defaults to -1.

property max_after_balance_size

Maximum relative size of the training data after balancing class counts (can be less than 1.0). Requires balance_classes.

Type: float, defaults to 5.0.

property max_confusion_matrix_size

[Deprecated] Maximum size (# classes) for confusion matrices to be printed in the Logs

Type: int, defaults to 20.

property max_iterations

Maximum number of iterations

Type: int, defaults to -1.

property max_runtime_secs

Maximum allowed runtime in seconds for model training. Use 0 to disable.

Type: float, defaults to 0.0.

property missing_values_handling

Handling of missing values. Either MeanImputation, Skip or PlugValues.

Type: Literal["mean_imputation", "skip", "plug_values"], defaults to "mean_imputation".

property nfolds

Number of folds for K-fold cross-validation (0 to disable or >= 2).

Type: int, defaults to 0.

property nlambdas

Number of lambdas to be used in a search. Default indicates: If alpha is zero, with lambda search set to True, the value of nlamdas is set to 30 (fewer lambdas are needed for ridge regression) otherwise it is set to 100.

Type: int, defaults to -1.

property non_negative

Restrict coefficients (not intercept) to be non-negative

Type: bool, defaults to False.

property num_knots

Number of knots for gam predictors. If specified, must specify one for each gam predictor. For monotone I-splines, mininum = 2, for cs spline, minimum = 3. For thin plate, minimum is size of polynomial basis + 2.

Type: List[int].

property obj_reg

Likelihood divider in objective value computation, default is 1/nobs

Type: float, defaults to -1.0.

property objective_epsilon

Converge if objective value changes less than this. Default indicates: If lambda_search is set to True the value of objective_epsilon is set to .0001. If the lambda_search is set to False and lambda is equal to zero, the value of objective_epsilon is set to .000001, for any other value of lambda the default value of objective_epsilon is set to .0001.

Type: float, defaults to -1.0.

property offset_column

Offset column. This will be added to the combination of columns before applying the link function.

Type: str.

property plug_values

Plug Values (a single row frame containing values that will be used to impute missing values of the training/validation frame, use with conjunction missing_values_handling = PlugValues)

Type: Union[None, str, H2OFrame].

property prior

Prior probability for y==1. To be used only for logistic regression iff the data has been sampled and the mean of response does not reflect reality.

Type: float, defaults to -1.0.

property remove_collinear_columns

In case of linearly dependent columns, remove some of the dependent columns

Type: bool, defaults to False.

property response_column

Response variable column.

Type: str.

property scale

Smoothing parameter for gam predictors. If specified, must be of the same length as gam_columns

Type: List[float].

property scale_tp_penalty_mat

Scale penalty matrix for tp (thin plate) smoothers as in R

Type: bool, defaults to False.

property score_each_iteration

Whether to score during each iteration of model training.

Type: bool, defaults to False.

scoring_history()[source]

Retrieve Model Score History.

Returns

The score history as an H2OTwoDimTable or a Pandas DataFrame.

property seed

Seed for pseudo random number generator (if applicable)

Type: int, defaults to -1.

property solver

AUTO will set the solver based on given data and the other parameters. IRLSM is fast on on problems with small number of predictors and for lambda-search with L1 penalty, L_BFGS scales better for datasets with many columns.

Type: Literal["auto", "irlsm", "l_bfgs", "coordinate_descent_naive", "coordinate_descent", "gradient_descent_lh", "gradient_descent_sqerr"], defaults to "auto".

property spline_orders

Order of I-splines or NBSplineTypeI M-splines used for gam predictors. If specified, must be the same size as gam_columns. For I-splines, the spline_orders will be the same as the polynomials used to generate the splines. For M-splines, the polynomials used to generate the splines will be spline_order-1. Values for bs=0 or 1 will be ignored.

Type: List[int].

property splines_non_negative

Valid for I-spline (bs=2) only. True if the I-splines are monotonically increasing (and monotonically non- decreasing) and False if the I-splines are monotonically decreasing (and monotonically non-increasing). If specified, must be the same size as gam_columns. Values for other spline types will be ignored. Default to true.

Type: List[bool].

property standardize

Standardize numeric columns to have zero mean and unit variance

Type: bool, defaults to False.

property standardize_tp_gam_cols

standardize tp (thin plate) predictor columns

Type: bool, defaults to False.

property startval

double array to initialize coefficients for GAM.

Type: List[float].

property stopping_metric

Metric to use for early stopping (AUTO: logloss for classification, deviance for regression and anomaly_score for Isolation Forest). Note that custom and custom_increasing can only be used in GBM and DRF with the Python client.

Type: Literal["auto", "deviance", "logloss", "mse", "rmse", "mae", "rmsle", "auc", "aucpr", "lift_top_group", "misclassification", "mean_per_class_error", "custom", "custom_increasing"], defaults to "auto".

property stopping_rounds

Early stopping based on convergence of stopping_metric. Stop if simple moving average of length k of the stopping_metric does not improve for k:=stopping_rounds scoring events (0 to disable)

Type: int, defaults to 0.

property stopping_tolerance

Relative tolerance for metric-based stopping criterion (stop if relative improvement is not at least this much)

Type: float, defaults to 0.001.

property theta

Theta

Type: float, defaults to 0.0.

property training_frame

Id of the training data frame.

Type: Union[None, str, H2OFrame].

Tweedie link power

Type: float, defaults to 0.0.

property tweedie_variance_power

Tweedie variance power

Type: float, defaults to 0.0.

property validation_frame

Id of the validation data frame.

Type: Union[None, str, H2OFrame].

property weights_column

Column with observation weights. Giving some observation a weight of zero is equivalent to excluding it from the dataset; giving an observation a relative weight of 2 is equivalent to repeating that row twice. Negative weights are not allowed. Note: Weights are per-row observation weights and do not increase the size of the data frame. This is typically the number of times a row is repeated, but non-integer values are supported as well. During training, rows with higher weights matter more, due to the larger loss function pre-factor. If you set weight = 0 for a row, the returned prediction frame at that row is zero and this is incorrect. To get an accurate prediction, remove all rows with weight == 0.

Type: str.

H2OGradientBoostingEstimator

class h2o.estimators.gbm.H2OGradientBoostingEstimator(model_id=None, training_frame=None, validation_frame=None, nfolds=0, keep_cross_validation_models=True, keep_cross_validation_predictions=False, keep_cross_validation_fold_assignment=False, score_each_iteration=False, score_tree_interval=0, fold_assignment='auto', fold_column=None, response_column=None, ignored_columns=None, ignore_const_cols=True, offset_column=None, weights_column=None, balance_classes=False, class_sampling_factors=None, max_after_balance_size=5.0, max_confusion_matrix_size=20, ntrees=50, max_depth=5, min_rows=10.0, nbins=20, nbins_top_level=1024, nbins_cats=1024, r2_stopping=None, stopping_rounds=0, stopping_metric='auto', stopping_tolerance=0.001, max_runtime_secs=0.0, seed=-1, build_tree_one_node=False, learn_rate=0.1, learn_rate_annealing=1.0, distribution='auto', quantile_alpha=0.5, tweedie_power=1.5, huber_alpha=0.9, checkpoint=None, sample_rate=1.0, sample_rate_per_class=None, col_sample_rate=1.0, col_sample_rate_change_per_level=1.0, col_sample_rate_per_tree=1.0, min_split_improvement=1e-05, histogram_type='auto', max_abs_leafnode_pred=None, pred_noise_bandwidth=0.0, categorical_encoding='auto', calibrate_model=False, calibration_frame=None, calibration_method='auto', custom_metric_func=None, custom_distribution_func=None, export_checkpoints_dir=None, in_training_checkpoints_dir=None, in_training_checkpoints_tree_interval=1, monotone_constraints=None, check_constant_response=True, gainslift_bins=-1, auc_type='auto', interaction_constraints=None)[source]

Bases: h2o.estimators.estimator_base.H2OEstimator

Gradient Boosting Machine

Builds gradient boosted trees on a parsed data set, for regression or classification. The default distribution function will guess the model type based on the response column type. Otherwise, the response column must be an enum for “bernoulli” or “multinomial”, and numeric for all other distributions.

property auc_type

Set default multinomial AUC type.

Type: Literal["auto", "none", "macro_ovr", "weighted_ovr", "macro_ovo", "weighted_ovo"], defaults to "auto".

property balance_classes

Balance training data class counts via over/under-sampling (for imbalanced data).

Type: bool, defaults to False.

Examples

>>> covtype = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/covtype/covtype.20k.data")
>>> covtype[54] = covtype[54].asfactor()
>>> predictors = covtype.columns[0:54]
>>> response = 'C55'
>>> train, valid = covtype.split_frame(ratios=[.8], seed=1234)
>>> cov_gbm = H2OGradientBoostingEstimator(balance_classes=True,
...                                        seed=1234)
>>> cov_gbm.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cov_gbm.logloss(valid=True)
property build_tree_one_node

Run on one node only; no network overhead but fewer cpus used. Suitable for small datasets.

Type: bool, defaults to False.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_gbm = H2OGradientBoostingEstimator(build_tree_one_node=True,
...                                         seed=1234)
>>> cars_gbm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> cars_gbm.auc(valid=True)
property calibrate_model

Use Platt Scaling (default) or Isotonic Regression to calculate calibrated class probabilities. Calibration can provide more accurate estimates of class probabilities.

Type: bool, defaults to False.

Examples

>>> ecology = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/gbm_test/ecology_model.csv")
>>> ecology['Angaus'] = ecology['Angaus'].asfactor()
>>> response = 'Angaus'
>>> train, calib = ecology.split_frame(seed = 12354)
>>> predictors = ecology.columns[3:13]
>>> w = h2o.create_frame(binary_fraction=1,
...                      binary_ones_fraction=0.5,
...                      missing_fraction=0,
...                      rows=744, cols=1)
>>> w.set_names(["weight"])
>>> train = train.cbind(w)
>>> ecology_gbm = H2OGradientBoostingEstimator(ntrees=10,
...                                            max_depth=5,
...                                            min_rows=10,
...                                            learn_rate=0.1,
...                                            distribution="multinomial",
...                                            weights_column="weight",
...                                            calibrate_model=True,
...                                            calibration_frame=calib)
>>> ecology_gbm.train(x=predictors,
...                   y="Angaus",
...                   training_frame=train)
>>> ecology_gbm.auc()
property calibration_frame

Data for model calibration

Type: Union[None, str, H2OFrame].

Examples

>>> ecology = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/gbm_test/ecology_model.csv")
>>> ecology['Angaus'] = ecology['Angaus'].asfactor()
>>> response = 'Angaus'
>>> predictors = ecology.columns[3:13]
>>> train, calib = ecology.split_frame(seed=12354)
>>> w = h2o.create_frame(binary_fraction=1,
...                      binary_ones_fraction=0.5,
...                      missing_fraction=0,
...                      rows=744,cols=1)
>>> w.set_names(["weight"])
>>> train = train.cbind(w)
>>> ecology_gbm = H2OGradientBoostingEstimator(ntrees=10,
...                                            max_depth=5,
...                                            min_rows=10,
...                                            learn_rate=0.1,
...                                            distribution="multinomial",
...                                            calibrate_model=True,
...                                            calibration_frame=calib)
>>> ecology_gbm.train(x=predictors,
...                   y="Angaus",
...                   training_frame=train,
...                   weights_column="weight")
>>> ecology_gbm.auc()
property calibration_method

Calibration method to use

Type: Literal["auto", "platt_scaling", "isotonic_regression"], defaults to "auto".

property categorical_encoding

Encoding scheme for categorical features

Type: Literal["auto", "enum", "one_hot_internal", "one_hot_explicit", "binary", "eigen", "label_encoder", "sort_by_response", "enum_limited"], defaults to "auto".

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Year"] = airlines["Year"].asfactor()
>>> airlines["Month"] = airlines["Month"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> train, valid = airlines.split_frame(ratios=[.8], seed=1234)
>>> airlines_gbm = H2OGradientBoostingEstimator(categorical_encoding="labelencoder",
...                                             seed=1234)
>>> airlines_gbm.train(x=predictors,
...                    y=response,
...                    training_frame=train,
...                    validation_frame=valid)
>>> airlines_gbm.auc(valid=True)
property check_constant_response

Check if response column is constant. If enabled, then an exception is thrown if the response column is a constant value.If disabled, then model will train regardless of the response column being a constant value or not.

Type: bool, defaults to True.

Examples

>>> train = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/iris/iris_train.csv")
>>> train["constantCol"] = 1
>>> my_gbm = H2OGradientBoostingEstimator(check_constant_response=False)
>>> my_gbm.train(x=list(range(1,5)),
...              y="constantCol",
...              training_frame=train)
property checkpoint

Model checkpoint to resume training with.

Type: Union[None, str, H2OEstimator].

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_gbm = H2OGradientBoostingEstimator(ntrees=1,
...                                         seed=1234)
>>> cars_gbm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> print(cars_gbm.auc(valid=True))
>>> print("Number of trees built for cars_gbm model:", cars_gbm.ntrees)
>>> cars_gbm_continued = H2OGradientBoostingEstimator(checkpoint=cars_gbm.model_id,
...                                                   ntrees=50,
...                                                   seed=1234)
>>> cars_gbm_continued.train(x=predictors,
...                          y=response,
...                          training_frame=train,
...                          validation_frame=valid)
>>> cars_gbm_continued.auc(valid=True)
>>> print("Number of trees built for cars_gbm model:",cars_gbm_continued.ntrees) 
property class_sampling_factors

Desired over/under-sampling ratios per class (in lexicographic order). If not specified, sampling factors will be automatically computed to obtain class balance during training. Requires balance_classes.

Type: List[float].

Examples

>>> covtype = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/covtype/covtype.20k.data")
>>> covtype[54] = covtype[54].asfactor()
>>> predictors = covtype.columns[0:54]
>>> response = 'C55'
>>> train, valid = covtype.split_frame(ratios=[.8], seed=1234)
>>> sample_factors = [1., 0.5, 1., 1., 1., 1., 1.]
>>> cov_gbm = H2OGradientBoostingEstimator(balance_classes=True,
...                                        class_sampling_factors=sample_factors,
...                                        seed=1234)
>>> cov_gbm.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cov_gbm.logloss(valid=True)
property col_sample_rate

Column sample rate (from 0.0 to 1.0)

Type: float, defaults to 1.0.

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Year"] = airlines["Year"].asfactor()
>>> airlines["Month"] = airlines["Month"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> train, valid = airlines.split_frame(ratios=[.8], seed=1234)
>>> airlines_gbm = H2OGradientBoostingEstimator(col_sample_rate=.7,
...                                             seed=1234)
>>> airlines_gbm.train(x=predictors,
...                    y=response,
...                    training_frame=train,
...                    validation_frame=valid)
>>> airlines_gbm.auc(valid=True)
property col_sample_rate_change_per_level

Relative change of the column sampling rate for every level (must be > 0.0 and <= 2.0)

Type: float, defaults to 1.0.

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Year"] = airlines["Year"].asfactor()
>>> airlines["Month"] = airlines["Month"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> train, valid = airlines.split_frame(ratios=[.8], seed=1234)
>>> airlines_gbm = H2OGradientBoostingEstimator(col_sample_rate_change_per_level=.9,
...                                             seed=1234)
>>> airlines_gbm.train(x=predictors,
...                    y=response,
...                    training_frame=train,
...                    validation_frame=valid)
>>> airlines_gbm.auc(valid=True)
property col_sample_rate_per_tree

Column sample rate per tree (from 0.0 to 1.0)

Type: float, defaults to 1.0.

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Year"] = airlines["Year"].asfactor()
>>> airlines["Month"] = airlines["Month"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> train, valid = airlines.split_frame(ratios=[.8], seed=1234)
>>> airlines_gbm = H2OGradientBoostingEstimator(col_sample_rate_per_tree=.7,
...                                             seed=1234)
>>> airlines_gbm.train(x=predictors,
...                    y=response,
...                    training_frame=train,
...                    validation_frame=valid)
>>> airlines_gbm.auc(valid=True)
property custom_distribution_func

Reference to custom distribution, format: language:keyName=funcName

Type: str.

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Year"] = airlines["Year"].asfactor()
>>> airlines["Month"] = airlines["Month"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> train, valid = airlines.split_frame(ratios=[.8], seed=1234)
>>> airlines_gbm = H2OGradientBoostingEstimator(ntrees=3,
...                                             max_depth=5,
...                                             distribution="bernoulli",
...                                             seed=1234)
>>> airlines_gbm.train(x=predictors,
...                    y=response,
...                    training_frame=train,
...                    validation_frame valid)
>>> from h2o.utils.distributions import CustomDistributionBernoulli
>>> custom_distribution_bernoulli = h2o.upload_custom_distribution(CustomDistributionBernoulli,
...                                                                func_name="custom_bernoulli",
...                                                                func_file="custom_bernoulli.py")
>>> airlines_gbm_custom = H2OGradientBoostingEstimator(ntrees=3,
...                                                    max_depth=5,
...                                                    distribution="custom",
...                                                    custom_distribution_func=custom_distribution_bernoulli,
...                                                    seed=1235)
>>> airlines_gbm_custom.train(x=predictors,
...                           y=response,
...                           training_frame=train,
...                           validation_frame=valid)
>>> airlines_gbm.auc()
property custom_metric_func

Reference to custom evaluation function, format: language:keyName=funcName

Type: str.

property distribution

Distribution function

Type: Literal["auto", "bernoulli", "quasibinomial", "multinomial", "gaussian", "poisson", "gamma", "tweedie", "laplace", "quantile", "huber", "custom"], defaults to "auto".

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> response = "cylinders"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_gbm = H2OGradientBoostingEstimator(distribution="poisson",
...                                         seed=1234)
>>> cars_gbm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> cars_gbm.mse(valid=True)
property export_checkpoints_dir

Automatically export generated models to this directory.

Type: str.

Examples

>>> airlines = h2o.import_file("http://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip", destination_frame="air.hex")
>>> predictors = ["DayofMonth", "DayOfWeek"]
>>> response = "IsDepDelayed"
>>> hyper_parameters = {'ntrees': [5,10]}
>>> search_crit = {'strategy': "RandomDiscrete",
...                'max_models': 5,
...                'seed': 1234,
...                'stopping_rounds': 3,
...                'stopping_metric': "AUTO",
...                'stopping_tolerance': 1e-2}
>>> checkpoints_dir = tempfile.mkdtemp()
>>> air_grid = H2OGridSearch(H2OGradientBoostingEstimator,
...                          hyper_params=hyper_parameters,
...                          search_criteria=search_crit)
>>> air_grid.train(x=predictors,
...                y=response,
...                training_frame=airlines,
...                distribution="bernoulli",
...                learn_rate=0.1,
...                max_depth=3,
...                export_checkpoints_dir=checkpoints_dir)
>>> len(listdir(checkpoints_dir))
property fold_assignment

Cross-validation fold assignment scheme, if fold_column is not specified. The ‘Stratified’ option will stratify the folds based on the response variable, for classification problems.

Type: Literal["auto", "random", "modulo", "stratified"], defaults to "auto".

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> assignment_type = "Random"
>>> cars_gbm = H2OGradientBoostingEstimator(fold_assignment=assignment_type,
...                                         nfolds=5,
...                                         seed=1234)
>>> cars_gbm.train(x=predictors, y=response, training_frame=cars)
>>> cars_gbm.auc(xval=True)
property fold_column

Column with cross-validation fold index assignment per observation.

Type: str.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> fold_numbers = cars.kfold_column(n_folds=5,
...                                  seed=1234)
>>> fold_numbers.set_names(["fold_numbers"])
>>> cars = cars.cbind(fold_numbers)
>>> cars_gbm = H2OGradientBoostingEstimator(seed=1234)
>>> cars_gbm.train(x=predictors,
...                y=response,
...                training_frame=cars,
...                fold_column="fold_numbers")
>>> cars_gbm.auc(xval=True)
property gainslift_bins

Gains/Lift table number of bins. 0 means disabled.. Default value -1 means automatic binning.

Type: int, defaults to -1.

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/testng/airlines_train.csv")
>>> model = H2OGradientBoostingEstimator(ntrees=1, gainslift_bins=20)
>>> model.train(x=["Origin", "Distance"],
...             y="IsDepDelayed",
...             training_frame=airlines)
>>> model.gains_lift()
property histogram_type

What type of histogram to use for finding optimal split points

Type: Literal["auto", "uniform_adaptive", "random", "quantiles_global", "round_robin", "uniform_robust"], defaults to "auto".

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Year"] = airlines["Year"].asfactor()
>>> airlines["Month"] = airlines["Month"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> train, valid = airlines.split_frame(ratios=[.8], seed=1234)
>>> airlines_gbm = H2OGradientBoostingEstimator(histogram_type="UniformAdaptive",
...                                             seed=1234)
>>> airlines_gbm.train(x=predictors,
...                    y=response,
...                    training_frame=train,
...                    validation_frame=valid)
>>> airlines_gbm.auc(valid=True)
property huber_alpha

Desired quantile for Huber/M-regression (threshold between quadratic and linear loss, must be between 0 and 1).

Type: float, defaults to 0.9.

Examples

>>> insurance = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/glm_test/insurance.csv")
>>> predictors = insurance.columns[0:4]
>>> response = 'Claims'
>>> insurance['Group'] = insurance['Group'].asfactor()
>>> insurance['Age'] = insurance['Age'].asfactor()
>>> train, valid = insurance.split_frame(ratios=[.8], seed=1234)
>>> insurance_gbm = H2OGradientBoostingEstimator(distribution="huber",
...                                              huber_alpha=0.9,
...                                              seed=1234)
>>> insurance_gbm.train(x=predictors,
...                     y=response,
...                     training_frame=train,
...                     validation_frame=valid)
>>> insurance_gbm.mse(valid=True)
property ignore_const_cols

Ignore constant columns.

Type: bool, defaults to True.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> cars["const_1"] = 6
>>> cars["const_2"] = 7
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_gbm = H2OGradientBoostingEstimator(seed=1234,
...                                         ignore_const_cols=True)
>>> cars_gbm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> cars_gbm.auc(valid=True)
property ignored_columns

Names of columns to ignore for training.

Type: List[str].

property in_training_checkpoints_dir

Create checkpoints into defined directory while training process is still running. In case of cluster shutdown, this checkpoint can be used to restart training.

Type: str.

property in_training_checkpoints_tree_interval

Checkpoint the model after every so many trees. Parameter is used only when in_training_checkpoints_dir is defined

Type: int, defaults to 1.

property interaction_constraints

A set of allowed column interactions.

Type: List[List[str]].

property keep_cross_validation_fold_assignment

Whether to keep the cross-validation fold assignment.

Type: bool, defaults to False.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> folds = 5
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_gbm = H2OGradientBoostingEstimator(keep_cross_validation_fold_assignment=True,
...                                         nfolds=5,
...                                         seed=1234)
>>> cars_gbm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> cars_gbm.auc()
property keep_cross_validation_models

Whether to keep the cross-validation models.

Type: bool, defaults to True.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> folds = 5
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_gbm = H2OGradientBoostingEstimator(keep_cross_validation_models=True,
...                                         nfolds=5,
...                                         seed=1234)
>>> cars_gbm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> cars_gbm.auc()
property keep_cross_validation_predictions

Whether to keep the predictions of the cross-validation models.

Type: bool, defaults to False.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> folds = 5
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_gbm = H2OGradientBoostingEstimator(keep_cross_validation_predictions=True,
...                                         nfolds=5,
...                                         seed=1234)
>>> cars_gbm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> cars_gbm.auc()
property learn_rate

Learning rate (from 0.0 to 1.0)

Type: float, defaults to 0.1.

Examples

>>> titanic = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/gbm_test/titanic.csv")
>>> titanic['survived'] = titanic['survived'].asfactor()
>>> predictors = titanic.columns
>>> del predictors[1:3]
>>> response = 'survived'
>>> train, valid = titanic.split_frame(ratios=[.8], seed=1234)
>>> titanic_gbm = H2OGradientBoostingEstimator(ntrees=10000,
...                                            learn_rate=0.01,
...                                            stopping_rounds=5,
...                                            stopping_metric="AUC",
...                                            stopping_tolerance=1e-4,
...                                            seed=1234)
>>> titanic_gbm.train(x=predictors,
...                   y=response,
...                   training_frame=train,
...                   validation_frame=valid)
>>> titanic_gbm.auc(valid=True)
property learn_rate_annealing

Scale the learning rate by this factor after each tree (e.g., 0.99 or 0.999)

Type: float, defaults to 1.0.

Examples

>>> titanic = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/gbm_test/titanic.csv")
>>> titanic['survived'] = titanic['survived'].asfactor()
>>> predictors = titanic.columns
>>> del predictors[1:3]
>>> response = 'survived'
>>> train, valid = titanic.split_frame(ratios=[.8], seed=1234)
>>> titanic_gbm = H2OGradientBoostingEstimator(ntrees=10000,
...                                            learn_rate=0.05,
...                                            learn_rate_annealing=.9,
...                                            stopping_rounds=5,
...                                            stopping_metric="AUC",
...                                            stopping_tolerance=1e-4,
...                                            seed=1234)
>>> titanic_gbm.train(x=predictors,
...                   y=response,
...                   training_frame=train,
...                   validation_frame=valid)
>>> titanic_gbm.auc(valid=True)
property max_abs_leafnode_pred

Maximum absolute value of a leaf node prediction

Type: float, defaults to .

Examples

>>> covtype = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/covtype/covtype.20k.data")
>>> covtype[54] = covtype[54].asfactor()
>>> predictors = covtype.columns[0:54]
>>> response = 'C55'
>>> train, valid = covtype.split_frame(ratios=[.8], seed=1234)
>>> cov_gbm = H2OGradientBoostingEstimator(max_abs_leafnode_pred=2,
...                                        seed=1234)
>>> cov_gbm.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cov_gbm.logloss(valid=True)
property max_after_balance_size

Maximum relative size of the training data after balancing class counts (can be less than 1.0). Requires balance_classes.

Type: float, defaults to 5.0.

Examples

>>> covtype = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/covtype/covtype.20k.data")
>>> covtype[54] = covtype[54].asfactor()
>>> predictors = covtype.columns[0:54]
>>> response = 'C55'
>>> train, valid = covtype.split_frame(ratios=[.8], seed=1234)
>>> max = .85
>>> cov_gbm = H2OGradientBoostingEstimator(balance_classes=True,
...                                        max_after_balance_size=max,
...                                        seed=1234)
>>> cov_gbm.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cov_gbm.logloss(valid=True)
property max_confusion_matrix_size

[Deprecated] Maximum size (# classes) for confusion matrices to be printed in the Logs

Type: int, defaults to 20.

property max_depth

Maximum tree depth (0 for unlimited).

Type: int, defaults to 5.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_gbm = H2OGradientBoostingEstimator(ntrees=100,
...                                         max_depth=2,
...                                         seed=1234)
>>> cars_gbm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> cars_gbm.auc(valid=True)
property max_runtime_secs

Maximum allowed runtime in seconds for model training. Use 0 to disable.

Type: float, defaults to 0.0.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_gbm = H2OGradientBoostingEstimator(max_runtime_secs=10,
...                                         ntrees=10000,
...                                         max_depth=10,
...                                         seed=1234)
>>> cars_gbm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> cars_gbm.auc(valid=True)
property min_rows

Fewest allowed (weighted) observations in a leaf.

Type: float, defaults to 10.0.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_gbm = H2OGradientBoostingEstimator(min_rows=16,
...                                         seed=1234)
>>> cars_gbm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> cars_gbm.auc(valid=True)
property min_split_improvement

Minimum relative improvement in squared error reduction for a split to happen

Type: float, defaults to 1e-05.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_gbm = H2OGradientBoostingEstimator(min_split_improvement=1e-3,
...                                         seed=1234)
>>> cars_gbm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> cars_gbm.auc(valid=True)
property monotone_constraints

A mapping representing monotonic constraints. Use +1 to enforce an increasing constraint and -1 to specify a decreasing constraint.

Type: dict.

Examples

>>> prostate_hex = h2o.import_file("http://s3.amazonaws.com/h2o-public-test-data/smalldata/prostate/prostate.csv.zip")
>>> prostate_hex["CAPSULE"] = prostate_hex["CAPSULE"].asfactor()
>>> response = "CAPSULE"
>>> seed = 42
>>> monotone_constraints = {"AGE":1}
>>> gbm_model = H2OGradientBoostingEstimator(seed=seed,
...                                          monotone_constraints=monotone_constraints)
>>> gbm_model.train(y=response,
...                 ignored_columns=["ID"],
...                 training_frame=prostate_hex)
>>> gbm_model.scoring_history()
property nbins

For numerical columns (real/int), build a histogram of (at least) this many bins, then split at the best point

Type: int, defaults to 20.

Examples

>>> eeg = h2o.import_file("https://h2o-public-test-data.s3.amazonaws.com/smalldata/eeg/eeg_eyestate.csv")
>>> eeg['eyeDetection'] = eeg['eyeDetection'].asfactor()
>>> predictors = eeg.columns[:-1]
>>> response = 'eyeDetection'
>>> train, valid = eeg.split_frame(ratios=[.8], seed=1234)
>>> bin_num = [16, 32, 64, 128, 256, 512]
>>> label = ["16", "32", "64", "128", "256", "512"]
>>> for key, num in enumerate(bin_num):
...     eeg_gbm = H2OGradientBoostingEstimator(nbins=num, seed=1234)
...     eeg_gbm.train(x=predictors,
...                   y=response,
...                   training_frame=train,
...                   validation_frame=valid)
...     print(label[key], 'training score', eeg_gbm.auc(train=True)) 
...     print(label[key], 'validation score', eeg_gbm.auc(valid=True))
property nbins_cats

For categorical columns (factors), build a histogram of this many bins, then split at the best point. Higher values can lead to more overfitting.

Type: int, defaults to 1024.

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Year"] = airlines["Year"].asfactor()
>>> airlines["Month"] = airlines["Month"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> train, valid = airlines.split_frame(ratios=[.8], seed=1234)
>>> bin_num = [8, 16, 32, 64, 128, 256, 512, 1024, 2048, 4096]
>>> label = ["8", "16", "32", "64", "128", "256", "512", "1024", "2048", "4096"]
>>> for key, num in enumerate(bin_num):
...     airlines_gbm = H2OGradientBoostingEstimator(nbins_cats=num, seed=1234)
...     airlines_gbm.train(x=predictors,
...                        y=response,
...                        training_frame=train,
...                        validation_frame=valid)
...     print(label[key], 'training score', airlines_gbm.auc(train=True))
...     print(label[key], 'validation score', airlines_gbm.auc(valid=True))
property nbins_top_level

For numerical columns (real/int), build a histogram of (at most) this many bins at the root level, then decrease by factor of two per level

Type: int, defaults to 1024.

Examples

>>> eeg = h2o.import_file("https://h2o-public-test-data.s3.amazonaws.com/smalldata/eeg/eeg_eyestate.csv")
>>> eeg['eyeDetection'] = eeg['eyeDetection'].asfactor()
>>> predictors = eeg.columns[:-1]
>>> response = 'eyeDetection'
>>> train, valid = eeg.split_frame(ratios=[.8], seed=1234)
>>> bin_num = [32, 64, 128, 256, 512, 1024, 2048, 4096]
>>> label = ["32", "64", "128", "256", "512", "1024", "2048", "4096"]
>>> for key, num in enumerate(bin_num):
...     eeg_gbm = H2OGradientBoostingEstimator(nbins_top_level=num, seed=1234)
...     eeg_gbm.train(x=predictors,
...                   y=response,
...                   training_frame=train,
...                   validation_frame=valid)
...     print(label[key], 'training score', eeg_gbm.auc(train=True)) 
...     print(label[key], 'validation score', eeg_gbm.auc(valid=True))
property nfolds

Number of folds for K-fold cross-validation (0 to disable or >= 2).

Type: int, defaults to 0.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> folds = 5
>>> cars_gbm = H2OGradientBoostingEstimator(nfolds=folds,
...                                         seed=1234
>>> cars_gbm.train(x=predictors,
...                y=response,
...                training_frame=cars)
>>> cars_gbm.auc()
property ntrees

Number of trees.

Type: int, defaults to 50.

Examples

>>> titanic = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/gbm_test/titanic.csv")
>>> titanic['survived'] = titanic['survived'].asfactor()
>>> predictors = titanic.columns
>>> del predictors[1:3]
>>> response = 'survived'
>>> train, valid = titanic.split_frame(ratios=[.8], seed=1234)
>>> tree_num = [20, 50, 80, 110, 140, 170, 200]
>>> label = ["20", "50", "80", "110", "140", "170", "200"]
>>> for key, num in enumerate(tree_num):
...     titanic_gbm = H2OGradientBoostingEstimator(ntrees=num,
...                                                seed=1234)
...     titanic_gbm.train(x=predictors,
...                       y=response,
...                       training_frame=train,
...                       validation_frame=valid)
...     print(label[key], 'training score', titanic_gbm.auc(train=True))
...     print(label[key], 'validation score', titanic_gbm.auc(valid=True))
property offset_column

Offset column. This will be added to the combination of columns before applying the link function.

Type: str.

Examples

>>> boston = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/gbm_test/BostonHousing.csv")
>>> predictors = boston.columns[:-1]
>>> response = "medv"
>>> boston['chas'] = boston['chas'].asfactor()
>>> boston["offset"] = boston["medv"].log()
>>> train, valid = boston.split_frame(ratios=[.8], seed=1234)
>>> boston_gbm = H2OGradientBoostingEstimator(offset_column="offset",
...                                           seed=1234)
>>> boston_gbm.train(x=predictors,
...                  y=response,
...                  training_frame=train,
...                  validation_frame=valid)
>>> boston_gbm.mse(valid=True)
property pred_noise_bandwidth

Bandwidth (sigma) of Gaussian multiplicative noise ~N(1,sigma) for tree node predictions

Type: float, defaults to 0.0.

Examples

>>> titanic = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/gbm_test/titanic.csv")
>>> titanic['survived'] = titanic['survived'].asfactor()
>>> predictors = titanic.columns
>>> del predictors[1:3]
>>> response = 'survived'
>>> train, valid = titanic.split_frame(ratios=[.8], seed=1234)
>>> titanic_gbm = H2OGradientBoostingEstimator(pred_noise_bandwidth=0.1,
...                                            seed=1234)
>>> titanic_gbm.train(x=predictors,
...                   y=response,
...                   training_frame=train,
...                   validation_frame=valid)
>>> titanic_gbm.auc(valid = True)
property quantile_alpha

Desired quantile for Quantile regression, must be between 0 and 1.

Type: float, defaults to 0.5.

Examples

>>> boston = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/gbm_test/BostonHousing.csv")
>>> predictors = boston.columns[:-1]
>>> response = "medv"
>>> boston['chas'] = boston['chas'].asfactor()
>>> train, valid = boston.split_frame(ratios=[.8], seed=1234)
>>> boston_gbm = H2OGradientBoostingEstimator(distribution="quantile",
...                                           quantile_alpha=.8,
...                                           seed=1234)
>>> boston_gbm.train(x=predictors,
...                  y=response,
...                  training_frame=train,
...                  validation_frame=valid)
>>> boston_gbm.mse(valid=True)
property r2_stopping

r2_stopping is no longer supported and will be ignored if set - please use stopping_rounds, stopping_metric and stopping_tolerance instead. Previous version of H2O would stop making trees when the R^2 metric equals or exceeds this

Type: float, defaults to .

property response_column

Response variable column.

Type: str.

property sample_rate

Row sample rate per tree (from 0.0 to 1.0)

Type: float, defaults to 1.0.

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Month"] = airlines["Month"].asfactor()                             >>> airlines["Year"]= airlines["Year"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> train, valid = airlines.split_frame(ratios=[.8], seed=1234)
>>> airlines_gbm = H2OGradientBoostingEstimator(sample_rate=.7,
...                                             seed=1234)
>>> airlines_gbm.train(x=predictors,
...                    y=response,
...                    training_frame=train,
...                    validation_frame=valid)
>>> airlines_gbm.auc(valid=True)
property sample_rate_per_class

A list of row sample rates per class (relative fraction for each class, from 0.0 to 1.0), for each tree

Type: List[float].

Examples

>>> covtype = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/covtype/covtype.20k.data")
>>> covtype[54] = covtype[54].asfactor()
>>> predictors = covtype.columns[0:54]
>>> response = 'C55'
>>> train, valid = covtype.split_frame(ratios=[.8], seed=1234)
>>> rate_per_class_list = [1, .4, 1, 1, 1, 1, 1]
>>> cov_gbm = H2OGradientBoostingEstimator(sample_rate_per_class=rate_per_class_list,
...                                        seed=1234)
>>> cov_gbm.train(x=predictors,
...               y=response,
...               training_frame=train,
...               validation_frame=valid)
>>> cov_gbm.logloss(valid=True)
property score_each_iteration

Whether to score during each iteration of model training.

Type: bool, defaults to False.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8],
...                                 seed=1234)
>>> cars_gbm = H2OGradientBoostingEstimator(score_each_iteration=True,
...                                         ntrees=55,
...                                         seed=1234)
>>> cars_gbm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> cars_gbm.scoring_history()
property score_tree_interval

Score the model after every so many trees. Disabled if set to 0.

Type: int, defaults to 0.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8],
...                                 seed=1234)
>>> cars_gbm = H2OGradientBoostingEstimator(score_tree_interval=True,
...                                         ntrees=55,
...                                         seed=1234)
>>> cars_gbm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> cars_gbm.scoring_history()
property seed

Seed for pseudo random number generator (if applicable)

Type: int, defaults to -1.

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Year"] = airlines["Year"].asfactor()
>>> airlines["Month"] = airlines["Month"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> train, valid = airlines.split_frame(ratios=[.8], seed=1234)
>>> gbm_w_seed_1 = H2OGradientBoostingEstimator(col_sample_rate=.7,
...                                             seed=1234)
>>> gbm_w_seed_1.train(x=predictors,
...                    y=response,
...                    training_frame=train,
...                    validation_frame=valid)
>>> print('auc for the 1st model built with a seed:', gbm_w_seed_1.auc(valid=True))
property stopping_metric

Metric to use for early stopping (AUTO: logloss for classification, deviance for regression and anomaly_score for Isolation Forest). Note that custom and custom_increasing can only be used in GBM and DRF with the Python client.

Type: Literal["auto", "deviance", "logloss", "mse", "rmse", "mae", "rmsle", "auc", "aucpr", "lift_top_group", "misclassification", "mean_per_class_error", "custom", "custom_increasing"], defaults to "auto".

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Year"] = airlines["Year"].asfactor()
>>> airlines["Month"] = airlines["Month"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> train, valid = airlines.split_frame(ratios=[.8], seed=1234)
>>> airlines_gbm = H2OGradientBoostingEstimator(stopping_metric="auc",
...                                             stopping_rounds=3,
...                                             stopping_tolerance=1e-2,
...                                             seed=1234)
>>> airlines_gbm.train(x=predictors,
...                    y=response,
...                    training_frame=train,
...                    validation_frame=valid)
>>> airlines_gbm.auc(valid=True)
property stopping_rounds

Early stopping based on convergence of stopping_metric. Stop if simple moving average of length k of the stopping_metric does not improve for k:=stopping_rounds scoring events (0 to disable)

Type: int, defaults to 0.

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Year"] = airlines["Year"].asfactor()
>>> airlines["Month"] = airlines["Month"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> train, valid = airlines.split_frame(ratios=[.8], seed=1234)
>>> airlines_gbm = H2OGradientBoostingEstimator(stopping_metric="auc",
...                                             stopping_rounds=3,
...                                             stopping_tolerance=1e-2,
...                                             seed=1234)
>>> airlines_gbm.train(x=predictors,
...                    y=response,
...                    training_frame=train,
...                    validation_frame=valid)
>>> airlines_gbm.auc(valid=True)
property stopping_tolerance

Relative tolerance for metric-based stopping criterion (stop if relative improvement is not at least this much)

Type: float, defaults to 0.001.

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Year"] = airlines["Year"].asfactor()
>>> airlines["Month"] = airlines["Month"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> train, valid= airlines.split_frame(ratios=[.8], seed=1234)
>>> airlines_gbm = H2OGradientBoostingEstimator(stopping_metric="auc",
...                                             stopping_rounds=3,
...                                             stopping_tolerance=1e-2,
...                                             seed=1234)
>>> airlines_gbm.train(x=predictors,
...                    y=response,
...                    training_frame=train,
...                    validation_frame=valid)
>>> airlines_gbm.auc(valid=True)
property training_frame

Id of the training data frame.

Type: Union[None, str, H2OFrame].

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_gbm = H2OGradientBoostingEstimator(seed=1234)
>>> cars_gbm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> cars_gbm.auc(valid=True)
property tweedie_power

Tweedie power for Tweedie regression, must be between 1 and 2.

Type: float, defaults to 1.5.

Examples

>>> insurance = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/glm_test/insurance.csv")
>>> predictors = insurance.columns[0:4]
>>> response = 'Claims'
>>> insurance['Group'] = insurance['Group'].asfactor()
>>> insurance['Age'] = insurance['Age'].asfactor()
>>> train, valid = insurance.split_frame(ratios=[.8], seed=1234)
>>> insurance_gbm = H2OGradientBoostingEstimator(distribution="tweedie",
...                                              tweedie_power=1.2,
...                                              seed=1234)
>>> insurance_gbm.train(x=predictors,
...                     y=response,
...                     training_frame=train,
...                     validation_frame=valid)
>>> insurance_gbm.mse(valid=True)
property validation_frame

Id of the validation data frame.

Type: Union[None, str, H2OFrame].

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_gbm = H2OGradientBoostingEstimator(seed=1234)
>>> cars_gbm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> cars_gbm.auc(valid=True)
property weights_column

Column with observation weights. Giving some observation a weight of zero is equivalent to excluding it from the dataset; giving an observation a relative weight of 2 is equivalent to repeating that row twice. Negative weights are not allowed. Note: Weights are per-row observation weights and do not increase the size of the data frame. This is typically the number of times a row is repeated, but non-integer values are supported as well. During training, rows with higher weights matter more, due to the larger loss function pre-factor. If you set weight = 0 for a row, the returned prediction frame at that row is zero and this is incorrect. To get an accurate prediction, remove all rows with weight == 0.

Type: str.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_gbm = H2OGradientBoostingEstimator(seed=1234)
>>> cars_gbm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid,
...                weights_column="weight")
>>> cars_gbm.auc(valid=True)

H2OGeneralizedLinearEstimator

class h2o.estimators.glm.H2OGeneralizedLinearEstimator(model_id=None, training_frame=None, validation_frame=None, nfolds=0, checkpoint=None, export_checkpoints_dir=None, seed=-1, keep_cross_validation_models=True, keep_cross_validation_predictions=False, keep_cross_validation_fold_assignment=False, fold_assignment='auto', fold_column=None, response_column=None, ignored_columns=None, random_columns=None, ignore_const_cols=True, score_each_iteration=False, score_iteration_interval=-1, offset_column=None, weights_column=None, family='auto', rand_family=None, tweedie_variance_power=0.0, tweedie_link_power=1.0, theta=1e-10, solver='auto', alpha=None, lambda_=None, lambda_search=False, early_stopping=True, nlambdas=-1, standardize=True, missing_values_handling='mean_imputation', plug_values=None, compute_p_values=False, dispersion_parameter_method='pearson', init_dispersion_parameter=1.0, remove_collinear_columns=False, intercept=True, non_negative=False, max_iterations=-1, objective_epsilon=-1.0, beta_epsilon=0.0001, gradient_epsilon=-1.0, link='family_default', rand_link=None, startval=None, calc_like=False, HGLM=False, prior=-1.0, cold_start=False, lambda_min_ratio=-1.0, beta_constraints=None, max_active_predictors=-1, interactions=None, interaction_pairs=None, obj_reg=-1.0, stopping_rounds=0, stopping_metric='auto', stopping_tolerance=0.001, balance_classes=False, class_sampling_factors=None, max_after_balance_size=5.0, max_confusion_matrix_size=20, max_runtime_secs=0.0, custom_metric_func=None, generate_scoring_history=False, auc_type='auto', dispersion_epsilon=0.0001, tweedie_epsilon=8e-17, max_iterations_dispersion=3000, build_null_model=False, fix_dispersion_parameter=False, generate_variable_inflation_factors=False, fix_tweedie_variance_power=True, dispersion_learning_rate=0.5)[source]

Bases: h2o.estimators.estimator_base.H2OEstimator

Generalized Linear Modeling

Fits a generalized linear model, specified by a response variable, a set of predictors, and a description of the error distribution.

A subclass of ModelBase is returned. The specific subclass depends on the machine learning task at hand (if it’s binomial classification, then an H2OBinomialModel is returned, if it’s regression then a H2ORegressionModel is returned). The default print-out of the models is shown, but further GLM-specific information can be queried out of the object. Upon completion of the GLM, the resulting object has coefficients, normalized coefficients, residual/null deviance, aic, and a host of model metrics including MSE, AUC (for logistic regression), degrees of freedom, and confusion matrices.

property HGLM

If set to true, will return HGLM model. Otherwise, normal GLM model will be returned

Type: bool, defaults to False.

property Lambda

[Deprecated] Use lambda_ instead

property alpha

Distribution of regularization between the L1 (Lasso) and L2 (Ridge) penalties. A value of 1 for alpha represents Lasso regression, a value of 0 produces Ridge regression, and anything in between specifies the amount of mixing between the two. Default value of alpha is 0 when SOLVER = ‘L-BFGS’; 0.5 otherwise.

Type: List[float].

Examples

>>> boston = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/gbm_test/BostonHousing.csv")
>>> predictors = boston.columns[:-1]
>>> response = "medv"
>>> boston['chas'] = boston['chas'].asfactor()
>>> train, valid = boston.split_frame(ratios=[.8])
>>> boston_glm = H2OGeneralizedLinearEstimator(alpha=.25)
>>> boston_glm.train(x=predictors,
...                  y=response,
...                  training_frame=train,
...                  validation_frame=valid)
>>> print(boston_glm.mse(valid=True))
property auc_type

Set default multinomial AUC type.

Type: Literal["auto", "none", "macro_ovr", "weighted_ovr", "macro_ovo", "weighted_ovo"], defaults to "auto".

property balance_classes

Balance training data class counts via over/under-sampling (for imbalanced data).

Type: bool, defaults to False.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> predictors = ["displacement","power","weight","year"]
>>> response = "acceleration"
>>> train, valid = cars.split_frame(ratios=[.8])
>>> cars_glm = H2OGeneralizedLinearEstimator(balance_classes=True,
...                                          seed=1234)
>>> cars_glm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> cars_glm.mse()
property beta_constraints

Beta constraints

Type: Union[None, str, H2OFrame].

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> predictors = ["displacement","power","weight","year"]
>>> response = "acceleration"
>>> train, valid = cars.split_frame(ratios=[.8])
>>> n = len(predictors)
>>> constraints = h2o.H2OFrame({'names':predictors,
...                             'lower_bounds': [-1000]*n,
...                             'upper_bounds': [1000]*n,
...                             'beta_given': [1]*n,
...                             'rho': [0.2]*n})
>>> cars_glm = H2OGeneralizedLinearEstimator(standardize=True,
...                                          beta_constraints=constraints)
>>> cars_glm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> cars_glm.mse()
property beta_epsilon

Converge if beta changes less (using L-infinity norm) than beta esilon, ONLY applies to IRLSM solver

Type: float, defaults to 0.0001.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> predictors = ["displacement","power","weight","year"]
>>> response = "acceleration"
>>> train, valid = cars.split_frame(ratios=[.8])
>>> cars_glm = H2OGeneralizedLinearEstimator(beta_epsilon=1e-3)
>>> cars_glm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> cars_glm.mse()
property build_null_model

If set, will build a model with only the intercept. Default to false.

Type: bool, defaults to False.

property calc_like

if true, will return likelihood function value for HGLM.

Type: bool, defaults to False.

property checkpoint

Model checkpoint to resume training with.

Type: Union[None, str, H2OEstimator].

property class_sampling_factors

Desired over/under-sampling ratios per class (in lexicographic order). If not specified, sampling factors will be automatically computed to obtain class balance during training. Requires balance_classes.

Type: List[float].

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> predictors = ["displacement","power","weight","year"]
>>> response = "acceleration"
>>> train, valid = cars.split_frame(ratios=[.8])
>>> sample_factors = [1., 0.5, 1., 1., 1., 1., 1.]
>>> cars_glm = H2OGeneralizedLinearEstimator(balance_classes=True,
...                                          class_sampling_factors=sample_factors,
...                                          seed=1234)
>>> cars_glm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> cars_glm.mse()
property cold_start

Only applicable to multiple alpha/lambda values. If false, build the next model for next set of alpha/lambda values starting from the values provided by current model. If true will start GLM model from scratch.

Type: bool, defaults to False.

property compute_p_values

Request p-values computation, p-values work only with IRLSM solver and no regularization

Type: bool, defaults to False.

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Year"] = airlines["Year"].asfactor()
>>> airlines["Month"] = airlines["Month"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> train, valid= airlines.split_frame(ratios=[.8])
>>> airlines_glm = H2OGeneralizedLinearEstimator(family='binomial',
...                                              lambda_=0,
...                                              remove_collinear_columns=True,
...                                              compute_p_values=True)
>>> airlines_glm.train(x=predictors,
...                    y=response,
...                    training_frame=train,
...                    validation_frame=valid)
>>> airlines_glm.mse()
property custom_metric_func

Reference to custom evaluation function, format: language:keyName=funcName

Type: str.

property dispersion_epsilon

If changes in dispersion parameter estimation or loglikelihood value is smaller than dispersion_epsilon, will break out of the dispersion parameter estimation loop using maximum likelihood.

Type: float, defaults to 0.0001.

property dispersion_learning_rate

Dispersion learning rate is only valid for tweedie family dispersion parameter estimation using ml. It must be > 0. This controls how much the dispersion parameter estimate is to be changed when the calculated loglikelihood actually decreases with the new dispersion. In this case, instead of setting new dispersion = dispersion - change, we set new dispersion = dispersion + dispersion_learning_rate * change. Defaults to 0.5.

Type: float, defaults to 0.5.

property dispersion_parameter_method

Method used to estimate the dispersion parameter for Tweedie, Gamma and Negative Binomial only.

Type: Literal["deviance", "pearson", "ml"], defaults to "pearson".

property early_stopping

Stop early when there is no more relative improvement on train or validation (if provided)

Type: bool, defaults to True.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8])
>>> cars_glm = H2OGeneralizedLinearEstimator(family='binomial',
...                                          early_stopping=True)
>>> cars_glm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> cars_glm.auc(valid=True)
property export_checkpoints_dir

Automatically export generated models to this directory.

Type: str.

Examples

>>> import tempfile
>>> from os import listdir
>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> predictors = ["displacement","power","weight","year"]
>>> response = "acceleration"
>>> train, valid = cars.split_frame(ratios=[.8])
>>> checkpoints = tempfile.mkdtemp()
>>> cars_glm = H2OGeneralizedLinearEstimator(export_checkpoints_dir=checkpoints,
...                                          seed=1234)
>>> cars_glm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> cars_glm.mse()
>>> len(listdir(checkpoints_dir))
property family

Family. Use binomial for classification with logistic regression, others are for regression problems.

Type: Literal["auto", "gaussian", "binomial", "fractionalbinomial", "quasibinomial", "ordinal", "multinomial", "poisson", "gamma", "tweedie", "negativebinomial"], defaults to "auto".

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8])
>>> cars_glm = H2OGeneralizedLinearEstimator(family='binomial')
>>> cars_glm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> cars_glm.auc(valid = True)
property fix_dispersion_parameter

Only used for Tweedie, Gamma and Negative Binomial GLM. If set, will use the dispsersion parameter in init_dispersion_parameter as the standard error and use it to calculate the p-values. Default to false.

Type: bool, defaults to False.

property fix_tweedie_variance_power

If true, will fix tweedie variance power value to the value set in tweedie_variance_power.

Type: bool, defaults to True.

property fold_assignment

Cross-validation fold assignment scheme, if fold_column is not specified. The ‘Stratified’ option will stratify the folds based on the response variable, for classification problems.

Type: Literal["auto", "random", "modulo", "stratified"], defaults to "auto".

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> assignment_type = "Random"
>>> cars_gml = H2OGeneralizedLinearEstimator(fold_assignment=assignment_type,
...                                          nfolds=5,
...                                          family='binomial',
...                                          seed=1234)
>>> cars_glm.train(x=predictors,
...                y=response,
...                training_frame=cars)
>>> cars_glm.auc(train=True)
property fold_column

Column with cross-validation fold index assignment per observation.

Type: str.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> fold_numbers = cars.kfold_column(n_folds=5, seed=1234)
>>> fold_numbers.set_names(["fold_numbers"])
>>> cars = cars.cbind(fold_numbers)
>>> print(cars['fold_numbers'])
>>>  cars_glm = H2OGeneralizedLinearEstimator(seed=1234,
...                                           family="binomial")
>>> cars_glm.train(x=predictors,
...                y=response,
...                training_frame=cars,
...                fold_column="fold_numbers")
>>> cars_glm.auc(xval=True)
property generate_scoring_history

If set to true, will generate scoring history for GLM. This may significantly slow down the algo.

Type: bool, defaults to False.

property generate_variable_inflation_factors

if true, will generate variable inflation factors for numerical predictors. Default to false.

Type: bool, defaults to False.

Examples

>>> training_data = h2o.import_file("http://h2o-public-test-data.s3.amazonaws.com/smalldata/glm_test/gamma_dispersion_factor_9_10kRows.csv")
>>> predictors = ['abs.C1.','abs.C2.','abs.C3.','abs.C4.','abs.C5.']
>>> response = 'resp'
>>> vif_glm = H2OGeneralizedLinearEstimator(family="gamma",
...                                         lambda_=0,
...                                         generate_variable_inflation_factors=True,
...                                         fold_assignment="modulo",
...                                         nfolds=3,
...                                         keep_cross_validation_models=True)
>>> vif_glm.train(x=predictors, y=response, training_frame=training_data)
>>> vif_glm.get_variable_inflation_factors()
static getAlphaBest(model)[source]

Extract best alpha value found from glm model.

Parameters

model – source lambda search model

Examples

>>> d = h2o.import_file("http://s3.amazonaws.com/h2o-public-test-data/smalldata/prostate/prostate.csv")
>>> m = H2OGeneralizedLinearEstimator(family = 'binomial',
...                                   lambda_search = True,
...                                   solver = 'COORDINATE_DESCENT')
>>> m.train(training_frame = d,
...         x = [2,3,4,5,6,7,8],
...         y = 1)
>>> bestAlpha = H2OGeneralizedLinearEstimator.getAlphaBest(m)
>>> print("Best alpha found is {0}".format(bestAlpha))
static getGLMRegularizationPath(model)[source]

Extract full regularization path explored during lambda search from glm model.

Parameters

model – source lambda search model

Examples

>>> d = h2o.import_file("http://s3.amazonaws.com/h2o-public-test-data/smalldata/prostate/prostate.csv")
>>> m = H2OGeneralizedLinearEstimator(family = 'binomial',
...                                   lambda_search = True,
...                                   solver = 'COORDINATE_DESCENT')
>>> m.train(training_frame = d,
...         x = [2,3,4,5,6,7,8],
...         y = 1)
>>> r = H2OGeneralizedLinearEstimator.getGLMRegularizationPath(m)
>>> m2 = H2OGeneralizedLinearEstimator.makeGLMModel(model=m,
...                                                 coefs=r['coefficients'][10])
>>> dev1 = r['explained_deviance_train'][10]
>>> p = m2.model_performance(d)
>>> dev2 = 1-p.residual_deviance()/p.null_deviance()
>>> print(dev1, " =?= ", dev2)
static getLambdaBest(model)[source]

Extract best lambda value found from glm model.

Parameters

model – source lambda search model

Examples

>>> d = h2o.import_file("http://s3.amazonaws.com/h2o-public-test-data/smalldata/prostate/prostate.csv")
>>> m = H2OGeneralizedLinearEstimator(family = 'binomial',
...                                   lambda_search = True,
...                                   solver = 'COORDINATE_DESCENT')
>>> m.train(training_frame = d,
...         x = [2,3,4,5,6,7,8],
...         y = 1)
>>> bestLambda = H2OGeneralizedLinearEstimator.getLambdaBest(m)
>>> print("Best lambda found is {0}".format(bestLambda))
static getLambdaMax(model)[source]

Extract the maximum lambda value used during lambda search.

Parameters

model – source lambda search model

Examples

>>> d = h2o.import_file("http://s3.amazonaws.com/h2o-public-test-data/smalldata/prostate/prostate.csv")
>>> m = H2OGeneralizedLinearEstimator(family = 'binomial',
...                                   lambda_search = True,
...                                   solver = 'COORDINATE_DESCENT')
>>> m.train(training_frame = d,
...         x = [2,3,4,5,6,7,8],
...         y = 1)
>>> maxLambda = H2OGeneralizedLinearEstimator.getLambdaMax(m)
>>> print("Maximum lambda found is {0}".format(maxLambda))
static getLambdaMin(model)[source]

Extract the minimum lambda value calculated during lambda search from glm model. Note that due to early stop, this minimum lambda value may not be used in the actual lambda search.

Parameters

model – source lambda search model

Examples

>>> d = h2o.import_file("http://s3.amazonaws.com/h2o-public-test-data/smalldata/prostate/prostate.csv")
>>> m = H2OGeneralizedLinearEstimator(family = 'binomial',
...                                   lambda_search = True,
...                                   solver = 'COORDINATE_DESCENT')
>>> m.train(training_frame = d,
...         x = [2,3,4,5,6,7,8],
...         y = 1)
>>> minLambda = H2OGeneralizedLinearEstimator.getLambdaMin(m)
>>> print("Minimum lambda found is {0}".format(minLambda))
property gradient_epsilon

Converge if objective changes less (using L-infinity norm) than this, ONLY applies to L-BFGS solver. Default (of -1.0) indicates: If lambda_search is set to False and lambda is equal to zero, the default value of gradient_epsilon is equal to .000001, otherwise the default value is .0001. If lambda_search is set to True, the conditional values above are 1E-8 and 1E-6 respectively.

Type: float, defaults to -1.0.

Examples

>>> boston = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/gbm_test/BostonHousing.csv")
>>> predictors = boston.columns[:-1]
>>> response = "medv"
>>> boston['chas'] = boston['chas'].asfactor()
>>> train, valid = boston.split_frame(ratios=[.8])
>>> boston_glm = H2OGeneralizedLinearEstimator(gradient_epsilon=1e-3)
>>> boston_glm.train(x=predictors,
...                  y=response,
...                  training_frame=train,
...                  validation_frame=valid)
>>> boston_glm.mse()
property ignore_const_cols

Ignore constant columns.

Type: bool, defaults to True.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> cars["const_1"] = 6
>>> cars["const_2"] = 7
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_glm = H2OGeneralizedLinearEstimator(seed=1234,
...                                          ignore_const_cols=True,
...                                          family="binomial")
>>> cars_glm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> cars_glm.auc(valid=True)
property ignored_columns

Names of columns to ignore for training.

Type: List[str].

property init_dispersion_parameter

Only used for Tweedie, Gamma and Negative Binomial GLM. Store the initial value of dispersion parameter. If fix_dispersion_parameter is set, this value will be used in the calculation of p-values.Default to 1.0.

Type: float, defaults to 1.0.

property interaction_pairs

A list of pairwise (first order) column interactions.

Type: List[tuple].

Examples

>>> df = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> XY = [df.names[i-1] for i in [1,2,3,4,6,8,9,13,17,18,19,31]]
>>> interactions = [XY[i-1] for i in [5,7,9]]
>>> m = H2OGeneralizedLinearEstimator(lambda_search=True,
...                                   family="binomial",
...                                   interactions=interactions)
>>> m.train(x=XY[:len(XY)], y=XY[-1],training_frame=df)
>>> m._model_json['output']['coefficients_table']
>>> coef_m = m._model_json['output']['coefficients_table']
>>> interaction_pairs = [("CRSDepTime", "UniqueCarrier"),
...                      ("CRSDepTime", "Origin"),
...                      ("UniqueCarrier", "Origin")]
>>> mexp = H2OGeneralizedLinearEstimator(lambda_search=True,
...                                      family="binomial",
...                                      interaction_pairs=interaction_pairs)
>>> mexp.train(x=XY[:len(XY)], y=XY[-1],training_frame=df)
>>> mexp._model_json['output']['coefficients_table']
property interactions

A list of predictor column indices to interact. All pairwise combinations will be computed for the list.

Type: List[str].

Examples

>>> boston = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/gbm_test/BostonHousing.csv")
>>> predictors = boston.columns[:-1]
>>> response = "medv"
>>> boston['chas'] = boston['chas'].asfactor()
>>> train, valid = boston.split_frame(ratios=[.8])
>>> interactions_list = ['crim', 'dis']
>>> boston_glm = H2OGeneralizedLinearEstimator(interactions=interactions_list) 
>>> boston_glm.train(x=predictors,
...                  y=response,
...                  training_frame=train,
...                  validation_frame=valid)
>>> boston_glm.mse()
property intercept

Include constant term in the model

Type: bool, defaults to True.

Examples

>>> iris = h2o.import_file("http://h2o-public-test-data.s3.amazonaws.com/smalldata/iris/iris_wheader.csv")
>>> iris['class'] = iris['class'].asfactor()
>>> predictors = iris.columns[:-1]
>>> response = 'class'
>>> train, valid = iris.split_frame(ratios=[.8])
>>> iris_glm = H2OGeneralizedLinearEstimator(family='multinomial',
...                                          intercept=True)
>>> iris_glm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> iris_glm.logloss(valid=True)
property keep_cross_validation_fold_assignment

Whether to keep the cross-validation fold assignment.

Type: bool, defaults to False.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_glm = H2OGeneralizedLinearEstimator(keep_cross_validation_fold_assignment=True,
...                                          nfolds=5,
...                                          seed=1234,
...                                          family="binomial")
>>> cars_glm.train(x=predictors,
...                y=response,
...                training_frame=train)
>>> cars_glm.cross_validation_fold_assignment()
property keep_cross_validation_models

Whether to keep the cross-validation models.

Type: bool, defaults to True.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_glm = H2OGeneralizedLinearEstimator(keep_cross_validation_models=True,
...                                          nfolds=5,
...                                          seed=1234,
...                                          family="binomial")
>>> cars_glm.train(x=predictors,
...                y=response,
...                training_frame=train)
>>> cars_glm_cv_models = cars_glm.cross_validation_models()
>>> print(cars_glm.cross_validation_models())
property keep_cross_validation_predictions

Whether to keep the predictions of the cross-validation models.

Type: bool, defaults to False.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_glm = H2OGeneralizedLinearEstimator(keep_cross_validation_predictions=True,
...                                          nfolds=5,
...                                          seed=1234,
...                                          family="binomial")
>>> cars_glm.train(x=predictors,
...                y=response,
...                training_frame=train)
>>> cars_glm.cross_validation_predictions()
property lambda_

Regularization strength

Type: List[float].

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Year"] = airlines["Year"].asfactor()
>>> airlines["Month"] = airlines["Month"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> train, valid = airlines.split_frame(ratios=[.8])
>>> airlines_glm = H2OGeneralizedLinearEstimator(family='binomial',
...                                              lambda_=.0001)
>>> airlines_glm.train(x=predictors,
...                    y=response
...                    trainig_frame=train,
...                    validation_frame=valid)
>>> print(airlines_glm.auc(valid=True))
property lambda_min_ratio

Minimum lambda used in lambda search, specified as a ratio of lambda_max (the smallest lambda that drives all coefficients to zero). Default indicates: if the number of observations is greater than the number of variables, then lambda_min_ratio is set to 0.0001; if the number of observations is less than the number of variables, then lambda_min_ratio is set to 0.01.

Type: float, defaults to -1.0.

Examples

>>> boston = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/gbm_test/BostonHousing.csv")
>>> predictors = boston.columns[:-1]
>>> response = "medv"
>>> boston['chas'] = boston['chas'].asfactor()
>>> train, valid = boston.split_frame(ratios=[.8])
>>> boston_glm = H2OGeneralizedLinearEstimator(lambda_min_ratio=.0001)
>>> boston_glm.train(x=predictors,
...                  y=response,
...                  training_frame=train,
...                  validation_frame=valid)
>>> boston_glm.mse()

Use lambda search starting at lambda max, given lambda is then interpreted as lambda min

Type: bool, defaults to False.

Examples

>>> boston = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/gbm_test/BostonHousing.csv")
>>> predictors = boston.columns[:-1]
>>> response = "medv"
>>> boston['chas'] = boston['chas'].asfactor()
>>> train, valid = boston.split_frame(ratios=[.8])
>>> boston_glm = H2OGeneralizedLinearEstimator(lambda_search=True)
>>> boston_glm.train(x=predictors,
...                  y=response,
...                  training_frame=train,
...                  validation_frame=valid)
>>> print(boston_glm.mse(valid=True))

Link function.

Type: Literal["family_default", "identity", "logit", "log", "inverse", "tweedie", "ologit"], defaults to "family_default".

Examples

>>> iris = h2o.import_file("http://h2o-public-test-data.s3.amazonaws.com/smalldata/iris/iris_wheader.csv")
>>> iris['class'] = iris['class'].asfactor()
>>> predictors = iris.columns[:-1]
>>> response = 'class'
>>> train, valid = iris.split_frame(ratios=[.8])
>>> iris_glm = H2OGeneralizedLinearEstimator(family='multinomial',
...                                          link='family_default')
>>> iris_glm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> iris_glm.logloss()
static makeGLMModel(model, coefs, threshold=0.5)[source]

Create a custom GLM model using the given coefficients.

Needs to be passed source model trained on the dataset to extract the dataset information from.

Parameters
  • model – source model, used for extracting dataset information

  • coefs – dictionary containing model coefficients

  • threshold – (optional, only for binomial) decision threshold used for classification

Examples

>>> d = h2o.import_file("http://s3.amazonaws.com/h2o-public-test-data/smalldata/prostate/prostate.csv")
>>> m = H2OGeneralizedLinearEstimator(family='binomial',
...                                   lambda_search=True,
...                                   solver='COORDINATE_DESCENT')
>>> m.train(training_frame=d,
...         x=[2,3,4,5,6,7,8],
...         y=1)
>>> r = H2OGeneralizedLinearEstimator.getGLMRegularizationPath(m)
>>> m2 = H2OGeneralizedLinearEstimator.makeGLMModel(model=m,
...                                                 coefs=r['coefficients'][10])
>>> dev1 = r['explained_deviance_train'][10]
>>> p = m2.model_performance(d)
>>> dev2 = 1-p.residual_deviance()/p.null_deviance()
>>> print(dev1, " =?= ", dev2)
property max_active_predictors

Maximum number of active predictors during computation. Use as a stopping criterion to prevent expensive model building with many predictors. Default indicates: If the IRLSM solver is used, the value of max_active_predictors is set to 5000 otherwise it is set to 100000000.

Type: int, defaults to -1.

Examples

>>> higgs= h2o.import_file("https://h2o-public-test-data.s3.amazonaws.com/smalldata/testng/higgs_train_5k.csv")
>>> predictors = higgs.names
>>> predictors.remove('response')
>>> response = "response"
>>> train, valid = higgs.split_frame(ratios=[.8])
>>> higgs_glm = H2OGeneralizedLinearEstimator(family='binomial',
...                                           max_active_predictors=200)
>>> higgs_glm.train(x=predictors,
...                 y=response,
...                 training_frame=train,
...                 validation_frame=valid)
>>> higgs_glm.auc()
property max_after_balance_size

Maximum relative size of the training data after balancing class counts (can be less than 1.0). Requires balance_classes.

Type: float, defaults to 5.0.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> predictors = ["displacement","power","weight","year"]
>>> response = "acceleration"
>>> train, valid = cars.split_frame(ratios=[.8])
>>> max = .85
>>> cars_glm = H2OGeneralizedLinearEstimator(balance_classes=True,
...                                          max_after_balance_size=max,
...                                          seed=1234)
>>> cars_glm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> cars_glm.mse()
property max_confusion_matrix_size

[Deprecated] Maximum size (# classes) for confusion matrices to be printed in the Logs

Type: int, defaults to 20.

property max_iterations

Maximum number of iterations

Type: int, defaults to -1.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8])
>>> cars_glm = H2OGeneralizedLinearEstimator(family='binomial',
...                                          max_iterations=50)
>>> cars_glm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> cars_glm.mse()
property max_iterations_dispersion

Control the maximum number of iterations in the dispersion parameter estimation loop using maximum likelihood.

Type: int, defaults to 3000.

property max_runtime_secs

Maximum allowed runtime in seconds for model training. Use 0 to disable.

Type: float, defaults to 0.0.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8])
>>> cars_glm = H2OGeneralizedLinearEstimator(max_runtime_secs=10,
...                                          seed=1234) 
>>> cars_glm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> cars_glm.mse()
property missing_values_handling

Handling of missing values. Either MeanImputation, Skip or PlugValues.

Type: Literal["mean_imputation", "skip", "plug_values"], defaults to "mean_imputation".

Examples

>>> boston = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/gbm_test/BostonHousing.csv")
>>> predictors = boston.columns[:-1]
>>> response = "medv"
>>> boston['chas'] = boston['chas'].asfactor()
>>> boston.insert_missing_values()
>>> train, valid = boston.split_frame(ratios=[.8])
>>> boston_glm = H2OGeneralizedLinearEstimator(missing_values_handling="skip")
>>> boston_glm.train(x=predictors,
...                  y=response,
...                  training_frame=train,
...                  validation_frame=valid)
>>> boston_glm.mse()
property nfolds

Number of folds for K-fold cross-validation (0 to disable or >= 2).

Type: int, defaults to 0.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> folds = 5
>>> cars_glm = H2OGeneralizedLinearEstimator(nfolds=folds,
...                                          seed=1234,
...                                          family='binomial')
>>> cars_glm.train(x=predictors,
...                y=response,
...                training_frame=cars)
>>> cars_glm.auc(xval=True)
property nlambdas

Number of lambdas to be used in a search. Default indicates: If alpha is zero, with lambda search set to True, the value of nlamdas is set to 30 (fewer lambdas are needed for ridge regression) otherwise it is set to 100.

Type: int, defaults to -1.

Examples

>>> boston = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/gbm_test/BostonHousing.csv")
>>> predictors = boston.columns[:-1]
>>> response = "medv"
>>> boston['chas'] = boston['chas'].asfactor()
>>> train, valid = boston.split_frame(ratios=[.8])
>>> boston_glm = H2OGeneralizedLinearEstimator(lambda_search=True,
...                                            nlambdas=50)
>>> boston_glm.train(x=predictors,
...                  y=response,
...                  training_frame=train,
...                  validation_frame=valid)
>>> print(boston_glm.mse(valid=True))
property non_negative

Restrict coefficients (not intercept) to be non-negative

Type: bool, defaults to False.

Examples

>>> airlines = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Year"] = airlines["Year"].asfactor()
>>> airlines["Month"] = airlines["Month"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> train, valid= airlines.split_frame(ratios=[.8])
>>> airlines_glm = H2OGeneralizedLinearEstimator(family='binomial',
...                                              non_negative=True)
>>> airlines_glm.train(x=predictors,
...                    y=response,
...                    training_frame=train,
...                    validation_frame=valid)
>>> airlines_glm.auc()
property obj_reg

Likelihood divider in objective value computation, default (of -1.0) will set it to 1/nobs

Type: float, defaults to -1.0.

Examples

>>> df = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/bigdata/laptop/glm_ordinal_logit/ordinal_multinomial_training_set.csv")
>>> df["C11"] = df["C11"].asfactor()
>>> ordinal_fit = H2OGeneralizedLinearEstimator(family="ordinal",
...                                             alpha=1.0,
...                                             lambda_=0.000000001,
...                                             obj_reg=0.00001,
...                                             max_iterations=1000,
...                                             beta_epsilon=1e-8,
...                                             objective_epsilon=1e-10)
>>> ordinal_fit.train(x=list(range(0,10)),
...                   y="C11",
...                   training_frame=df)
>>> ordinal_fit.mse()
property objective_epsilon

Converge if objective value changes less than this. Default (of -1.0) indicates: If lambda_search is set to True the value of objective_epsilon is set to .0001. If the lambda_search is set to False and lambda is equal to zero, the value of objective_epsilon is set to .000001, for any other value of lambda the default value of objective_epsilon is set to .0001.

Type: float, defaults to -1.0.

Examples

>>> boston = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/gbm_test/BostonHousing.csv")
>>> predictors = boston.columns[:-1]
>>> response = "medv"
>>> boston['chas'] = boston['chas'].asfactor()
>>> train, valid = boston.split_frame(ratios=[.8])
>>> boston_glm = H2OGeneralizedLinearEstimator(objective_epsilon=1e-3)
>>> boston_glm.train(x=predictors,
...                  y=response,
...                  training_frame=train,
...                  validation_frame=valid)
>>> boston_glm.mse()
property offset_column

Offset column. This will be added to the combination of columns before applying the link function.

Type: str.

Examples

>>> boston = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/gbm_test/BostonHousing.csv")
>>> predictors = boston.columns[:-1]
>>> response = "medv"
>>> boston['chas'] = boston['chas'].asfactor()
>>> boston["offset"] = boston["medv"].log()
>>> train, valid = boston.split_frame(ratios=[.8], seed=1234)
>>> boston_glm = H2OGeneralizedLinearEstimator(offset_column="offset",
...                                            seed=1234)
>>> boston_glm.train(x=predictors,
...                  y=response,
...                  training_frame=train,
...                  validation_frame=valid)
>>> boston_glm.mse(valid=True)
property plug_values

Plug Values (a single row frame containing values that will be used to impute missing values of the training/validation frame, use with conjunction missing_values_handling = PlugValues)

Type: Union[None, str, H2OFrame].

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars = cars.drop(0)
>>> means = cars.mean()
>>> means = H2OFrame._expr(ExprNode("mean", cars, True, 0))
>>> glm_means = H2OGeneralizedLinearEstimator(seed=42)
>>> glm_means.train(training_frame=cars, y="cylinders")
>>> glm_plugs1 = H2OGeneralizedLinearEstimator(seed=42,
...                                            missing_values_handling="PlugValues",
...                                            plug_values=means)
>>> glm_plugs1.train(training_frame=cars, y="cylinders")
>>> glm_means.coef() == glm_plugs1.coef()
>>> not_means = 0.1 + (means * 0.5)
>>> glm_plugs2 = H2OGeneralizedLinearEstimator(seed=42,
...                                            missing_values_handling="PlugValues",
...                                            plug_values=not_means)
>>> glm_plugs2.train(training_frame=cars, y="cylinders")
>>> glm_means.coef() != glm_plugs2.coef()
property prior

Prior probability for y==1. To be used only for logistic regression iff the data has been sampled and the mean of response does not reflect reality.

Type: float, defaults to -1.0.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8])
>>> cars_glm1 = H2OGeneralizedLinearEstimator(family='binomial', prior=0.5)
>>> cars_glm1.train(x=predictors,
...                 y=response,
...                 training_frame=train,
...                 validation_frame=valid)
>>> cars_glm1.mse()
property rand_family

Random Component Family array. One for each random component. Only support gaussian for now.

Type: List[Literal["[gaussian]"]].

Link function array for random component in HGLM.

Type: List[Literal["[identity]", "[family_default]"]].

property random_columns

random columns indices for HGLM.

Type: List[int].

property remove_collinear_columns

In case of linearly dependent columns, remove some of the dependent columns

Type: bool, defaults to False.

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Year"] = airlines["Year"].asfactor()
>>> airlines["Month"] = airlines["Month"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> train, valid = airlines.split_frame(ratios=[.8])
>>> airlines_glm = H2OGeneralizedLinearEstimator(family='binomial',
...                                              lambda_=0,
...                                              remove_collinear_columns=True)
>>> airlines_glm.train(x=predictors,
...                    y=response,
...                    training_frame=train,
...                    validation_frame=valid)
>>> airlines_glm.auc()
property response_column

Response variable column.

Type: str.

property score_each_iteration

Whether to score during each iteration of model training.

Type: bool, defaults to False.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_glm = H2OGeneralizedLinearEstimator(score_each_iteration=True,
...                                          seed=1234,
...                                          family='binomial')
>>> cars_glm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> cars_glm.scoring_history()
property score_iteration_interval

Perform scoring for every score_iteration_interval iterations

Type: int, defaults to -1.

property seed

Seed for pseudo random number generator (if applicable)

Type: int, defaults to -1.

Examples

>>> airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")
>>> airlines["Year"] = airlines["Year"].asfactor()
>>> airlines["Month"] = airlines["Month"].asfactor()
>>> airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
>>> airlines["Cancelled"] = airlines["Cancelled"].asfactor()
>>> airlines['FlightNum'] = airlines['FlightNum'].asfactor()
>>> predictors = ["Origin", "Dest", "Year", "UniqueCarrier",
...               "DayOfWeek", "Month", "Distance", "FlightNum"]
>>> response = "IsDepDelayed"
>>> train, valid = airlines.split_frame(ratios=[.8], seed=1234)
>>> glm_w_seed = H2OGeneralizedLinearEstimator(family='binomial',
...                                            seed=1234)
>>> glm_w_seed.train(x=predictors,
...                  y=response,
...                  training_frame=train,
...                  validation_frame=valid)
>>> print(glm_w_seed_1.auc(valid=True))
property solver

AUTO will set the solver based on given data and the other parameters. IRLSM is fast on on problems with small number of predictors and for lambda-search with L1 penalty, L_BFGS scales better for datasets with many columns.

Type: Literal["auto", "irlsm", "l_bfgs", "coordinate_descent_naive", "coordinate_descent", "gradient_descent_lh", "gradient_descent_sqerr"], defaults to "auto".

Examples

>>> boston = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/gbm_test/BostonHousing.csv")
>>> predictors = boston.columns[:-1]
>>> response = "medv"
>>> boston['chas'] = boston['chas'].asfactor()
>>> train, valid = boston.split_frame(ratios=[.8])
>>> boston_glm = H2OGeneralizedLinearEstimator(solver='irlsm')
>>> boston_glm.train(x=predictors,
...                  y=response,
...                  training_frame=train,
...                  validation_frame=valid)
>>> print(boston_glm.mse(valid=True))
property standardize

Standardize numeric columns to have zero mean and unit variance

Type: bool, defaults to True.

Examples

>>> boston = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/gbm_test/BostonHousing.csv")
>>> predictors = boston.columns[:-1]
>>> response = "medv"
>>> boston['chas'] = boston['chas'].asfactor()
>>> train, valid = boston.split_frame(ratios=[.8])
>>> boston_glm = H2OGeneralizedLinearEstimator(standardize=True)
>>> boston_glm.train(x=predictors,
...                  y=response,
...                  training_frame=train,
...                  validation_frame=valid)
>>> boston_glm.mse()
property startval

double array to initialize fixed and random coefficients for HGLM, coefficients for GLM.

Type: List[float].

property stopping_metric

Metric to use for early stopping (AUTO: logloss for classification, deviance for regression and anomaly_score for Isolation Forest). Note that custom and custom_increasing can only be used in GBM and DRF with the Python client.

Type: Literal["auto", "deviance", "logloss", "mse", "rmse", "mae", "rmsle", "auc", "aucpr", "lift_top_group", "misclassification", "mean_per_class_error", "custom", "custom_increasing"], defaults to "auto".

property stopping_rounds

Early stopping based on convergence of stopping_metric. Stop if simple moving average of length k of the stopping_metric does not improve for k:=stopping_rounds scoring events (0 to disable)

Type: int, defaults to 0.

property stopping_tolerance

Relative tolerance for metric-based stopping criterion (stop if relative improvement is not at least this much)

Type: float, defaults to 0.001.

property theta

Theta

Type: float, defaults to 1e-10.

Examples

>>> h2o_df = h2o.import_file("http://h2o-public-test-data.s3.amazonaws.com/smalldata/glm_test/Motor_insurance_sweden.txt")
>>> predictors = ["Payment", "Insured", "Kilometres", "Zone", "Bonus", "Make"]
>>> response = "Claims"
>>> negativebinomial_fit = H2OGeneralizedLinearEstimator(family="negativebinomial",
...                                                      link="identity",
...                                                      theta=0.5)
>>> negativebinomial_fit.train(x=predictors,
...                            y=response,
...                            training_frame=h2o_df)
>>> negativebinomial_fit.scoring_history()
property training_frame

Id of the training data frame.

Type: Union[None, str, H2OFrame].

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8],
...                                 seed=1234)
>>> cars_glm = H2OGeneralizedLinearEstimator(seed=1234,
...                                          family='binomial')
>>> cars_glm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> cars_glm.auc(train=True)
property tweedie_epsilon

In estimating tweedie dispersion parameter using maximum likelihood, this is used to choose the lower and upper indices in the approximating of the infinite series summation.

Type: float, defaults to 8e-17.

Tweedie link power

Type: float, defaults to 1.0.

Examples

>>> auto = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/glm_test/auto.csv")
>>> predictors = auto.names
>>> predictors.remove('y')
>>> response = "y"
>>> train, valid = auto.split_frame(ratios=[.8])
>>> auto_glm = H2OGeneralizedLinearEstimator(family='tweedie',
...                                          tweedie_link_power=1)
>>> auto_glm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> print(auto_glm.mse(valid=True))
property tweedie_variance_power

Tweedie variance power

Type: float, defaults to 0.0.

Examples

>>> auto = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/glm_test/auto.csv")
>>> predictors = auto.names
>>> predictors.remove('y')
>>> response = "y"
>>> train, valid = auto.split_frame(ratios=[.8])
>>> auto_glm = H2OGeneralizedLinearEstimator(family='tweedie',
...                                          tweedie_variance_power=1)
>>> auto_glm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> print(auto_glm.mse(valid=True))
property validation_frame

Id of the validation data frame.

Type: Union[None, str, H2OFrame].

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","weight","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_glm = H2OGeneralizedLinearEstimator(seed=1234,
...                                          family='binomial')
>>> cars_glm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid)
>>> cars_glm.auc(valid=True)
property weights_column

Column with observation weights. Giving some observation a weight of zero is equivalent to excluding it from the dataset; giving an observation a relative weight of 2 is equivalent to repeating that row twice. Negative weights are not allowed. Note: Weights are per-row observation weights and do not increase the size of the data frame. This is typically the number of times a row is repeated, but non-integer values are supported as well. During training, rows with higher weights matter more, due to the larger loss function pre-factor. If you set weight = 0 for a row, the returned prediction frame at that row is zero and this is incorrect. To get an accurate prediction, remove all rows with weight == 0.

Type: str.

Examples

>>> cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")
>>> cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()
>>> predictors = ["displacement","power","acceleration","year"]
>>> response = "economy_20mpg"
>>> train, valid = cars.split_frame(ratios=[.8], seed=1234)
>>> cars_glm = H2OGeneralizedLinearEstimator(seed=1234,
...                                          family='binomial')
>>> cars_glm.train(x=predictors,
...                y=response,
...                training_frame=train,
...                validation_frame=valid,
...                weights_column="weight")
>>> cars_glm.auc(valid=True)

H2OInfogram

class h2o.estimators.infogram.H2OInfogram(model_id=None, training_frame=None, validation_frame=None, seed=-1, keep_cross_validation_models=True, keep_cross_validation_predictions=False, keep_cross_validation_fold_assignment=False, nfolds=0, fold_assignment='auto', fold_column=None, response_column=None, ignored_columns=None, ignore_const_cols=True, score_each_iteration=False, offset_column=None, weights_column=None, standardize=False, distribution='auto', plug_values=None, max_iterations=0, stopping_rounds=0, stopping_metric='auto', stopping_tolerance=0.001, balance_classes=False, class_sampling_factors=None, max_after_balance_size=5.0, max_runtime_secs=0.0, custom_metric_func=None, auc_type='auto', algorithm='auto', algorithm_params=None, protected_columns=None, total_information_threshold=-1.0, net_information_threshold=-1.0, relevance_index_threshold=-1.0, safety_index_threshold=-1.0, data_fraction=1.0, top_n_features=50)[source]

Bases: h2o.estimators.estimator_base.H2OEstimator

Information Diagram

The infogram is a graphical information-theoretic interpretability tool which allows the user to quickly spot the core, decision-making variables that uniquely and safely drive the response, in supervised classification problems. The infogram can significantly cut down the number of predictors needed to build a model by identifying only the most valuable, admissible features. When protected variables such as race or gender are present in the data, the admissibility of a variable is determined by a safety and relevancy index, and thus serves as a diagnostic tool for fairness. The safety of each feature can be quantified and variables that are unsafe will be conside